Distance measuring device, distance measuring method, program and depth map generating method

Abstract

Distance measuring device comprising: a first determination unit that determines whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; and a second determination unit that, if the first determination unit determines that a distance difference between the first pixel and the second pixel is greater than the first threshold, determines whether a confidence difference between the first pixel and the second pixel is greater than a second threshold, wherein, if the second determination unit determines that the confidence difference between the first pixel and the second pixel is greater than the second threshold, it confirms that the first pixel is a defective pixel.

Description

TECHNICAL AREA The present technology relates to a distance measuring device, a distance measuring method, a program, and a method for generating a depth map. In particular, the present technology relates to a distance measuring device, a distance measuring method, a program, and a method for generating a depth map, which are designed, for example, to detect false distance measurement results with high accuracy. BACKGROUND TECHNOLOGY In recent years, distance measuring devices, which measure distances to objects, have become increasingly smaller due to advances in semiconductor technology. As a result, distance measuring devices are now being integrated into mobile devices such as smartphones, which are small information processing devices or systems with communication capabilities. Examples of distance measuring devices (sensors) that measure distances to objects include a time-of-flight (TOF) sensor (see, for example, patent document 1). QUOTE LIST PATENT DOCUMENT Patent document 1: JP 2016 - 90 268 A US 2009 / 0016638 A1 deals with the detection of defective pixel elements of a CCD sensor. US patent 2008 / 0143856 A1 describes the determination of differences in the output values ​​of adjacent pixel elements, also for the detection of defective pixel elements. The subsequently published German patent DE 11 2019 000 317 T5 describes an image sensor device that incorporates a time-of-flight camera. German patent DE 10 2018 107 801 A1 discloses a time-of-flight camera for a time-of-flight camera system. Upon detection of a point light source, an image captured by the system can be corrected. SUMMARY OF THE INVENTION PROBLEMS TO BE SOLVED BY THE INVENTION If there is an incorrect distance measurement result, the incorrect distance measurement result is preferably detected with high accuracy, so that the accuracy of the distance measurement is improved as desired. The present technology was developed with such circumstances in mind and aims to enable accurate detection of incorrect distance measurement results. SOLUTIONS FOR THE PROBLEMS A first distance measuring device according to one aspect of the present technology comprises: a first determination unit that determines whether the depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; and a second determination unit that determines whether the confidence difference between the first pixel and the second pixel is greater than a second threshold if the first determination unit determines that the distance difference between the first pixel and the second pixel is greater than the first threshold. If the second determination unit determines that the confidence difference between the first pixel and the second pixel is greater than the second threshold, it is confirmed that the first pixel is a defective pixel. A second distance measuring device according to one aspect of the present technology comprises: a first determining unit that determines whether the depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; and a second determining unit that determines whether the reflectance difference between the first pixel and the second pixel is greater than a second threshold if the first determining unit determines that the distance difference between the first pixel and the second pixel is greater than the first threshold. If the second determining unit determines that the reflectance difference between the first pixel and the second pixel is greater than the second threshold, it is confirmed that the first pixel is a defective pixel. A first distance measurement method according to one aspect of the present technology is implemented by a distance measuring device that performs a distance measurement and comprises: determining whether the depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether the confidence difference between the first pixel and the second pixel is greater than a second threshold if it is determined that the distance difference between the first pixel and the second pixel is greater than the first threshold; and confirming that the first pixel is a defective pixel if it is determined that the confidence difference between the first pixel and the second pixel is greater than the second threshold. A second distance measurement method according to one aspect of the present technology is implemented by a distance measuring device that performs a distance measurement and comprises: determining whether the depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether the reflectance difference between the first pixel and the second pixel is greater than a second threshold if the distance difference between the first pixel and the second pixel is determined to be greater than the first threshold; and confirming that the first pixel is a defective pixel if the reflectance difference between the first pixel and the second pixel is determined to be greater than the second threshold. A first program according to one aspect of the present technology serves to cause a computer to perform a process comprising the steps of: determining whether the depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether the confidence difference between the first pixel and the second pixel is greater than a second threshold if the distance difference between the first pixel and the second pixel is determined to be greater than the first threshold; and confirming that the first pixel is a defective pixel if the confidence difference between the first pixel and the second pixel is determined to be greater than the second threshold. A second program, according to one aspect of the present technology, serves to cause a computer to perform a process comprising the steps of: determining whether the depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether the reflectance difference between the first pixel and the second pixel is greater than a second threshold if the distance difference between the first pixel and the second pixel is determined to be greater than the first threshold; and confirming that the first pixel is a defective pixel if the reflectance difference between the first pixel and the second pixel is determined to be greater than the second threshold. In the first distance measuring device, the first distance measuring method, and the first program according to any aspect of the present technology, if it is determined that the depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold, a further check is performed to determine whether the confidence difference between the first pixel and the second pixel is greater than a second threshold. If it is determined that the confidence difference between the first pixel and the second pixel is greater than the second threshold, it is then confirmed that the first pixel is a defective pixel. In the second distance measuring device, the second distance measuring method, and the second program, if it is determined that the depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold, a further check is performed to determine whether the reflectance difference between the first pixel and the second pixel is greater than a second threshold. If it is determined that the reflectance difference between the first pixel and the second pixel is greater than the second threshold, it is then confirmed that the first pixel is a defective pixel. An electronic device according to an aspect of the present technology not claimed herein comprises a sensor that performs a process using a machine-learned learning model on at least a portion of a first distance measurement information acquired by the sensor and outputs a second distance measurement information after a correction target pixel contained in the first distance measurement information has been corrected. The process comprises: a first process to correct the correction target pixel using the first distance measurement information containing the correction target pixel as input; and a second process to output the second distance measurement information. In the electronic device according to an aspect of the present technology not claimed herein, when a second distance measurement information is output after a correction target pixel contained in the first distance measurement information has been corrected by a process using a machine-learned learning model on at least part of the first distance measurement information acquired by the sensor, a first process is carried out to correct the correction target pixel using the first distance measurement information containing the correction target pixel as input, and a second process is carried out to output the second distance measurement information. A first method for generating a learning model according to an aspect of the present technology not claimed here comprises: acquiring training data containing a depth map which is a depth map containing a correction target pixel, wherein the position of the correction target pixel in the depth map has been identified; and generating a learning model using the training data, wherein the learning model receives an input of the depth map containing the correction target pixel and outputs position information about the correction target pixel contained in the depth map. In the first method, which generates a learning model according to an aspect of the present technology not claimed here, training data is acquired which contains a depth map containing a correction target pixel whose position in the depth map has been identified, and a learning model is generated using the training data, wherein the learning model receives an input of the depth map containing the correction target pixel and outputs position information about the correction target pixel contained in the depth map. A second method for generating a learning model, according to an aspect of the present technology not claimed here, comprises: acquiring training data containing a depth map including a correction target pixel and position information about the correction target pixel and a corrected depth map or position information about a corrected correction target pixel; and generating a learning model using the training data, wherein the learning model receives an input of the depth map containing the correction target pixel and the position information about the correction target pixel and outputs the corrected depth map or the position information about the corrected correction target pixel. In the second method, which generates a learning model according to one aspect of the present technology, training data containing a depth map including a correction target pixel and position information about the correction target pixel and a corrected depth map or position information about a corrected correction target pixel are acquired, and a learning model is generated using the training data, wherein the learning model receives an input of the depth map containing the correction target pixel and the position information about the correction target pixel and outputs the corrected depth map or the position information about the corrected correction target pixel. A depth map generating method according to one aspect of the present technology is implemented by a distance measuring device that performs a distance measurement and comprises: determining whether the distance difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether the confidence or reflectance difference between the first pixel and the second pixel is greater than a second threshold if the distance difference between the first pixel and the second pixel is determined to be greater than the first threshold; and confirming that the first pixel is a correction target pixel if the confidence or reflectance difference between the first pixel and the second pixel is determined to be greater than the second threshold.and the generation of a depth map in which the confirmed correction target pixel has been corrected. In a depth map generating method according to one aspect of the present technology, a check is performed to determine whether the distance difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; a check is performed to determine whether the confidence or reflectance difference between the first pixel and the second pixel is greater than a second threshold if the distance difference between the first pixel and the second pixel is determined to be greater than the first threshold; the first pixel is confirmed to be a correction target pixel if the confidence or distance difference between the first pixel and the second pixel is determined to be greater than the second threshold; and a depth map is generated in which the confirmed correction target pixel has been corrected. Note that the distance measuring devices and electronic devices or equipment may be independent devices or internal blocks within a facility. Furthermore, the program to be provided can be transmitted via a transmission medium or can be recorded on a recording medium. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a diagram illustrating the configuration of an embodiment of a distance measuring device for which the present technology is used. Fig. 2 is a diagram illustrating an exemplary configuration of a light-receiving unit. Fig. 3 is a diagram illustrating an exemplary configuration of a pixel. Fig. 4 is a graphical representation to explain a distribution of electrical charges in a pixel. Fig. 5 is a graphical representation illustrating an example of four types of received light, each time with a phase delay of 90 degrees. Fig. 6 is a graphical representation illustrating an example of detection signals during a detection period of received light with a phase delay of 0 degrees. Fig. 7 is a graphical representation illustrating an example of detection signals during a detection period of received light with a phase delay of 90 degrees.Figure 8 is a graphical representation showing an example of detection signals during a detection period of received light with a phase delay of 180 degrees. Figure 9 is a graphical representation showing an example of detection signals during a detection period of received light with a phase delay of 270 degrees. Figure 10 is a graphical representation to illustrate detection signals in a frame. Figure 11 is a graphical representation to illustrate the relationship between detection periods and detection signals. Figure 12 is a graphical representation to illustrate detection signals in a frame. Figure 13 is a graphical representation to illustrate detection signals in a frame. Figure 14 is a diagram to illustrate flying pixels. Figure 15 is a diagram to illustrate flying pixels. Figure 16 is a diagram to illustrate flying pixels.Figure 17 is a diagram to explain flying pixels. Figure 18 is a flowchart to explain a first process relating to the detection of flying pixels. Figure 19 is a diagram to explain the relationship between the process goal and the surrounding pixels. Figure 20 is a flowchart to explain a second process relating to the detection of flying pixels. Figure 21 is a diagram to explain the setting of a threshold. Figure 22 is a block diagram showing an exemplary configuration of an electronic device. Figure 23 is a block diagram showing an exemplary configuration of a personal computer. Figure 24 is a diagram showing an exemplary configuration of an embodiment of an electronic device for which the present technology is used. Figure 25 is a flowchart to explain the first processing of distance measurement information.Figure 26 is a flowchart illustrating a process for using a learning model. Figure 27 is a flowchart illustrating a process for correcting defective pixels. Figure 28 is a flowchart illustrating a secondary processing step involving distance measurement information. Figure 29 is a diagram showing an exemplary configuration of an embodiment of a learning device for which the present technology is used. Figure 30 is a diagram showing an example of a learning model. Figure 31 is a flowchart illustrating a learning process for generating the learning model. Figure 32 is a diagram showing an exemplary procedure for identifying defective pixels using an autocoder. Figure 33 is a flowchart illustrating a process for identifying defective pixels.Figure 34 is a diagram illustrating an exemplary procedure for correcting defective pixels using an autocoder. Figure 35 is a flowchart illustrating a process for correcting defective pixels. Figure 36 is a diagram illustrating an example of a learning model. Figure 37 is a flowchart illustrating a learning process for generating a learning model. Figure 38 is a diagram illustrating another exemplary configuration of an embodiment of an electronic device for which the present technology is used. Figure 39 is a flowchart illustrating a process for generating a corrected depth map. Figure 40 is a diagram illustrating an example of the processing unit that performs a machine learning process. Figure 41 is a perspective view of an exemplary external configuration of a sensor.Figure 42 is a diagram showing an exemplary configuration of a system containing devices for which the present technology is used. Figure 43 is a diagram showing an exemplary schematic configuration of a system for endoscopic surgery. Figure 44 is a block diagram showing an example of the functional configurations of a camera head and a CCU. Figure 45 is a block diagram showing an exemplary schematic configuration of a vehicle control system. Figure 46 is an explanatory diagram showing an example of installation positions for external information detectors and imaging units. MODES FOR EXECUTING THE INVENTION The following is a description of modes (referred to hereafter as embodiments) for implementing the present technology. The present technology can be used for light-receiving elements that form a distance measurement system which performs distance measurement, for example, by means of an indirect TOF method, and for an imaging device and the like that contains such light-receiving elements. For example, a distance measurement system for an in-vehicle system that measures the distance to a target object outside the vehicle, a gesture recognition system that measures the distance to a target object such as a user's hand and recognizes a user gesture based on the measurement result, and similar systems can be used. In this case, the result of the gesture recognition can be used, for example, to operate a vehicle navigation system or similar. <Beispielhafte Konfiguration einer Abstandsmessvorrichtung> Fig. 1 is a diagram showing an exemplary configuration of an embodiment of a distance measuring device for which the present technology is used. A distance measuring device 10 comprises a lens 11, a light receiving unit 12, a signal processing unit 13, a light emitting unit 14, a light emission control unit 15, and a filter unit 16. The distance measuring device 10 in Fig. 1 illuminates an object with light and receives light (reflected light), which is the light reflected from the object (irradiation light), in order to measure the distance to the object. The light emission system of the distance measuring device 10 consists of the light-emitting unit 14 and the light emission control unit 15. Within the light emission system, the light emission control unit 15, under the control of the signal processing unit 13, causes the light-emitting unit 14 to emit infrared (IR) light. An IR bandpass filter can be provided between the lens 11 and the light-receiving unit 12, and the light-emitting unit 14 can emit infrared light corresponding to the transmission wavelength band of the IR bandpass filter. The light-emitting unit 14 can be arranged inside the housing of the distance measuring device 10 or it can be arranged outside the housing of the distance measuring device 10. The light emission control unit 15 causes the light-emitting unit 14 to emit light at a predetermined frequency. The signal processing unit 13 acts as a calculation unit that calculates the distance (depth value) from the distance measuring device 10 to an object, for example, based on a detection signal (pixel data) provided by the light-receiving unit 12. The signal processing unit 13 generates a depth map in which the depth value (depth information) is stored as the pixel value of each pixel 50 (Fig. 2) of the light-receiving unit 12, and outputs the depth map to the filter unit 16. The signal processing unit 13 also calculates the confidence of the depth value calculated with respect to each pixel 50 of the light-receiving unit 12, generates a confidence map that stores the confidence (luminance information) as the pixel value of each pixel 50 of the light-receiving unit 12, and outputs the confidence map to the filter unit 16. <Konfiguration eines Bildgebungselements> Fig. 2 is a block diagram showing an exemplary configuration of the light-receiving unit 12. The light-receiving unit 12 can be an image sensor made of a complementary metal oxide semiconductor (CMOS). The light-receiving unit 12 comprises a pixel array unit 41, a vertical drive unit 42, a column processing unit 43, a horizontal drive unit 44, and a system control unit 45. The pixel array unit 41, the vertical drive unit 42, the column processing unit 43, the horizontal drive unit 44, and the system control unit 45 are formed on a semiconductor substrate (a chip) which is not shown in the drawing. In the pixel array unit 41, unit pixels (pixel 50 in Fig. 3, for example) with photoelectric conversion elements, which generate and store photoelectric charges corresponding to a quantity of incident light, are arranged two-dimensionally in a matrix. Note that in the description below, a photoelectric charge corresponding to a quantity of incident light can simply be referred to as "electric charge," and a unit pixel can simply be referred to as "pixel." In the matrix-like pixel array of the pixel array unit 41, pixel control lines 46 for the respective rows are formed in the transverse direction (the array direction of the pixels in the pixel rows) in the drawing, and vertical signal lines 47 for the respective columns are formed in the vertical direction (the array direction of the pixels in the pixel columns) in the drawing. One end of each pixel control line 46 is connected to the output end of the vertical control unit 42 corresponding to the respective rows. The vertical control unit 42 is a pixel control unit equipped with a shift register, an address decoder, and the like, which controls the respective pixels in the pixel array unit 41, for example, together or row by row. Pixel signals emitted by the respective unit pixels of a pixel row selected and scanned by the vertical control unit 42 are fed into the column processing unit 43 via the respective vertical signal lines 47. For the respective pixel columns of the pixel array unit 41, the column processing unit 43 performs predetermined signal processing on pixel signals emitted by the respective unit pixels of a selected row via the vertical signal lines 47 and temporarily holds the pixel signals undergoing signal processing. Specifically, the column processing unit 43 performs at least one noise reduction process, such as correlated double sampling (CDS), as part of its signal processing. This correlated double sampling performed by the column processing unit 43 removes reset noise and fixed-pattern noise unique to the pixels, such as threshold variation among gain transistors. Note that the column processing unit 43 may not only be designed to perform the noise reduction process, but also, for example, to perform analog-to-digital (AD) conversion and output signal levels as digital signals. The horizontal control unit 44 is equipped with a shift register, an address decoder, and the like, and sequentially selects the unit circuits according to the pixel columns of the column processing unit 43. As a result of this selective scanning by the horizontal control unit 44, the pixel signals that have undergone signal processing by the column processing unit 43 are sequentially output to the signal processing unit 48. The system control unit 45 is equipped with a time control generator or the like, which generates various time control signals, and performs a control operation for the vertical control unit 42, the column processing unit 43, the horizontal control unit 44 and the like on the basis of the various time control signals generated by the time control generator. In the matrix-like pixel array of the pixel array unit 41, the pixel control lines 46 are provided in the row direction in the respective pixel rows, and two vertical signal lines 47 are provided in the column direction in each pixel column. For example, the pixel control lines 46 transmit control signals to perform control when signals are read from the pixels. Note that in Fig. 2, each pixel control line 46 is shown as a wiring line, but is not necessarily a wiring line. One end of each pixel control line 46 is connected to the output end of the vertical control unit 42 corresponding to the respective row. <Struktur der Einheitspixel> Next, a specific structure of the unit pixels 50 arranged in a matrix in the pixel array unit 41 is described. A pixel 50 contains a photodiode 61 (hereinafter referred to as PD 61), which is a photoelectric conversion element, and is designed such that electrical charges generated in the PD 61 are distributed to taps 51-1 and 51-2. The electrical charge distributed to tap 51-1 is then read from a vertical signal line 47-1 and output as the detection signal SIG1. The electrical charge distributed to tap 51-2 is also read from a vertical signal line 47-2 and output as the detection signal SIG2. Tap 51-1 contains a transfer transistor 62-1, a floating diffusion region (FD) 63-1, a reset transistor 64, a gain transistor 65-1, and a selection transistor 66-1. Similarly, tap 51-2 contains a transfer transistor 62-2, an FD 63-2, the reset transistor 64, a gain transistor 65-2, and a selection transistor 66-2. Note that, as shown in Fig. 3, the reset transistor 64 can be used jointly by the FD 63-1 and the FD 63-2, or can be provided in both the FD 63-1 and the FD 63-2. If a reset transistor 64 is provided in both the FD 63-1 and the FD 63-2, the reset time for each can be controlled individually, thus enabling finer and more precise control. If a single reset transistor 64 is used for both the FD 63-1 and the FD 63-2, the reset time can be set to the same for both, simplifying the control and the circuit configuration. In the exemplary case described below, the reset transistor 64, which is common to the FD 63-1 and the FD 63-2, is used. Referring to Fig. 4, the distribution of electrical charges in a pixel 50 is now described. Here, the distribution means that the electrical charges accumulated in a pixel 50 (PD 61) are read at different times in order to perform a reading for the respective taps. As shown in Fig. 4, the light-emitting unit 14 emits irradiation light which is modulated (one cycle = Tp) in order to repeatedly switch irradiation on / off within the irradiation time, and reflected light is received by the PD 61 with a delay time Td depending on the distance to an object. A transfer control signal TRT_A controls the switching on / off of transfer transistor 62-1, and a transfer control signal TRT_B controls the switching on / off of transfer transistor 62-2. As shown in the drawing, while the transfer control signal TRT_A is in the same phase as the irradiation light, the transfer control signal TRT_B is in a phase that is the inverse of the transfer control signal TRT_A. Accordingly, while the transmission transistor 62-1 is switched on according to the transmission control signal TRT_A, electrical charges generated by the reflected light-receiving photodiode 61 are transferred to the FD area 63-1. Conversely, while the transmission transistor 62-2 is switched on according to the transmission control signal TRT_B, the electrical charges are transferred to the FD area 63-2. Consequently, during a predetermined period, while the irradiation light is emitted periodically with irradiation time T, the electrical charges transferred via the transmission transistor 62-1 are sequentially accumulated in the FD area 63-1, and the electrical charges transferred via the transmission transistor 62-2 are sequentially accumulated in the FD area 63-2. When the selection transistor 66-1 is switched on according to a selection signal SELm1 after the end of the period during which electrical charges are accumulated, the electrical charges accumulated in the FD area 63-1 are read out via the vertical signal line 47-1, and the detection signal A corresponding to the amount of charge is output by the light-receiving unit 12. Likewise, when the selection transistor 66-2 is switched on according to a selection signal SELm2, the electrical charges accumulated in the FD area 63-2 are read out via the vertical signal line 47-2, and the detection signal B corresponding to the amount of charge is output by the light-receiving unit 12. The electrical charges accumulated in FD area 63-1 are released when the reset transistor 64 is switched on according to a reset signal RST. Likewise, the electrical charges accumulated in FD area 63-2 are released when the reset transistor 64 is switched on according to the reset signal RST. As described above, pixel 50 can distribute the electrical charges generated by the photodiode 61, which receives the reflected light, to tap 51-1 and tap 51-2 according to the delay time Td and output the detection signal A and the detection signal B. Furthermore, the delay time Td corresponds to the time it takes for light emitted by the light-emitting unit 14 to travel to the object, be reflected by the object, and then travel to the light-receiving unit 12, which is the distance to the object. Consequently, the distance measuring device 10 can calculate the distance (depth) to the object based on the detection signal A and the detection signal B from the delay time Td. <Abstandsmessung mit der auf der Lichtempfangsseite verschobenen Phase> Referring to Fig. 5, Fig. 6, Fig. 7, Fig. 8 to Fig. 9, a case is described in which the receiving side receives light with a shifted phase and detects signals. As shown in Fig. 5, for example, four types of light reception are performed, each with a phase delay of 90 degrees. That is, based on a light reception with a phase delay of 0 degrees to receive light without any phase shift relative to that of the light-emitting side, light reception with a phase delay of 90 degrees, light reception with a phase delay of 180 degrees, and light reception with a phase delay of 270 degrees are performed, and one period (quad) is provided four times to detect the detection signal A and the detection signal B. That is, as shown in Fig. 6, for example, a detection period Q0 to detect reflected light by receiving light with a phase delay of 0 degrees, a detection period Q1 to detect reflected light by detecting light with a phase delay of 90 degrees, a detection period Q2 to detect reflected light by receiving light with a phase delay of 180 degrees, and a detection period Q3 to detect reflected light by receiving light with a phase delay of 270 degrees are continuously provided. Fig. 6 shows an example of the received light, the reflected light, the transmission control signal TRT_A, the transmission control signal TRT_B, the detection signal A, and the detection signal B during the detection period Q0. As shown in Fig. 6, the transmission control signal TRT_A of tap 51-1 is switched on at the same time (a phase delay of 0 degrees) as the start of emission of the incident light, and light reception (transmission) at tap 51-1 begins. At the time to switch off the transmission control signal TRT_A, the transmission control signal TRT_B is also switched on, and tap 51-2 begins receiving (transmitting) light. While light reception with a phase delay of 0 degrees is performed in this manner, electrical charges are distributed between tap 51-1 and tap 51-2, the amount of charge depending on the delay time TR, and the respective electrical charges are accumulated during one integration period. Subsequently, during a readout period, the electrical charges accumulated during the integration period are read out, and a detection signal A0 and a detection signal B0 are output during the detection period Q0. Fig. 7 shows an example of the received light, the reflected light, the transmission control signal TRT_A, the transmission control signal TRT_B, the detection signal A, and the detection signal B during the detection period Q1. As shown in Fig. 7, the transmission control signal TRT_A of tap 51-1 is switched on at a time with a phase delay of 90 degrees relative to the start of emission of the incident light, and light reception (transmission) at tap 51-1 begins. At the time to switch off the transmission control signal TRT_A, the transmission control signal TRT_B is also switched on, and tap 51-2 begins receiving (transmitting) light. While light reception with a 90-degree phase delay is performed in this manner, electrical charges are distributed between tap 51-1 and tap 51-2, the amount of charge depending on the delay time TR, and the respective electrical charges are accumulated during the integration period. Subsequently, during the readout period, the electrical charges accumulated during the integration period are read out, and a detection signal A90 and a detection signal B90 are output during the detection period Q1. Fig. 8 shows an example of the received light, the reflected light, the transmission control signal TRT_A, the transmission control signal TRT_B, the detection signal A, and the detection signal B during the detection period Q2. As shown in Fig. 8, the transmission control signal TRT_A of tap 51-1 is switched on at a time with a phase delay of 180 degrees relative to the start of emission of the incident light, and light reception (transmission) at tap 51-1 begins. At the time to switch off the transmission control signal TRT_A, the transmission control signal TRT_B is also switched on, and tap 51-2 begins receiving (transmitting) light. While light reception with a phase delay of 180 degrees is performed in this manner, electrical charges are distributed between tap 51-1 and tap 51-2, the amount of charge depending on the delay time TR, and the respective electrical charges are accumulated during the integration period. Subsequently, during a readout period, the electrical charges accumulated during the integration period are read out, and a detection signal A180 and a detection signal B180 are output during the detection period Q2. Fig. 9 shows an example of the received light, the reflected light, the transmission control signal TRT_A, the transmission control signal TRT_B, the detection signal A, and the detection signal B during the detection period Q3. As shown in Fig. 9, the transmission control signal TRT_A of tap 51-1 is switched on at a time with a phase delay of 270 degrees relative to the start of emission of the incident light, and light reception (transmission) at tap 51-1 begins. At the time to switch off the transmission control signal TRT_A, the transmission control signal TRT_B is also switched on, and tap 51-2 begins receiving (transmitting) light. While light reception with a phase delay of 270 degrees is performed in this manner, electrical charges are distributed between tap 51-1 and tap 51-2, the amount of charge depending on the delay time TR, and the respective electrical charges are accumulated during the integration period. Subsequently, during the readout period, the electrical charges accumulated during the integration period are read out, and a detection signal A270 and a detection signal B270 are output during the detection period Q3. As described above, detection signals A0 and B0 are detected via light reception with a phase delay of 0 degrees during detection period Q0, and detection signals A90 and B90 are detected via light reception with a phase delay of 90 degrees during detection period Q1. Similarly, detection signals A180 and B180 are detected via light reception with a phase delay of 180 degrees during detection period Q2, and detection signals A270 and B270 are detected via light reception with a phase delay of 270 degrees during detection period Q3. On the light-emitting side, despite a phase delay at the beginning of light reception, there are no phase delays. In other words, the emission of irradiated light always begins at the same time. If light reception with four phase delays is carried out by the two taps 51 as described above and the distance to a predetermined object is measured, a process is carried out by the signal processing unit 13 ( Fig. 1) on the basis of the signals received during the four detection periods. <Berechnung eines Abstands> As shown in Fig. 10, the detection period of a frame comprises a detection period Q0, a detection period Q1, a detection period Q2, and a detection period Q3. A detection signal A0 and a detection signal B0 are acquired during detection period Q0, and a detection signal A90 and a detection signal B90 are acquired during detection period Q1. A detection signal A180 and a detection signal B180 are also acquired during detection period Q2, and a detection signal A270 and a detection signal B270 are acquired during detection period Q3. The signal processing unit 13 (Fig. 1) uses these detection signals to calculate a phase difference θ according to expression (1) shown below, a distance D according to expression (2) shown below, and a confidence level according to expression (3) shown below. [Mathematical Formula 1] [Mathematical Formula 2] [Mathematical Formula 3] In expression (1), I represents the value obtained by subtracting the value C180, obtained by subtracting detection signal B180 from detection signal A180, from the value C0, obtained by subtracting detection signal B0 from detection signal A0. Q represents the value obtained by subtracting the value C270, obtained by subtracting detection signal B270 from detection signal A270, from the value C90, obtained by subtracting detection signal B90 from detection signal A90. The phase difference θ is calculated by determining the arctangent of (Q / I). In expression (2), C represents the speed of light and Tp represents the pulse width. The delay time Td can be obtained based on the phase difference θ, and the distance D to the object can be calculated based on the delay time Td. Expression (3) is an equation for calculating the value representing the confidence of the calculated distance. The confidence c is obtained by calculating the square root of the value obtained by adding the square of I to the square of Q. Note that calculating the confidence c is not an essential element in calculating the distance D and can be omitted. Alternatively, the confidence can be calculated according to an expression other than expression (3). For example, the sum of the absolute values ​​of I and Q can be defined as the confidence c. The explanation continues here assuming that the confidence is calculated according to expression (3), but a case where the confidence is calculated according to another expression is also an application of the present technology. The description below provides an example of a case where the distance to a predetermined object is calculated using the two Taps 51 sensors and irradiation with four phase differences (referred to hereafter as "2-Tap 4-Phase"). However, the technology can also be used in a case where the distance to a predetermined object is measured using the two Taps 51 sensors and irradiation with two phase differences, or in a case where the distance to a predetermined object is measured using one Tap 51 sensor and irradiation with four phase differences. <2-Tap 2-Phase> A case in which the distance to a predetermined object is measured using the two taps 51 with irradiated light with two phase differences or by receiving light with two phase differences (hereinafter referred to as "2-tap 2-phase") is briefly described below. The explanation continues here, taking as an example a case in which light reception with two phase differences is performed to measure the distance to a predetermined object. Fig. 11 is a diagram showing the exposure times of the four phases 0 degrees, 90 degrees, 180 degrees and 270 degrees, with the beginnings of the phase differences aligned for ease of understanding. In practice, as shown in Fig. 10, imaging is performed during a detection period Q0 to detect a detection signal A0 and a detection signal B0, imaging is performed during a detection period Q1 to detect a detection signal A90 and a detection signal B90, imaging is performed during a detection period Q2 to detect a detection signal A180 and a detection signal B180, and imaging is performed during a detection period Q3 to detect a detection signal A270 and a detection signal B270. Fig. 11 shows this imaging process, which is performed sequentially in the time direction, while the beginnings of the detection periods are aligned in the vertical direction. From the beginning of detection period Q0, an exposure is performed to image the detection signal A0, followed by an exposure to image the detection signal B0. From the moment the phase is shifted by 90 degrees relative to the start of the detection period Q1, an exposure is performed to image the detection signal A90, followed by an exposure to image the detection signal B90. From the point in time when the phase is shifted by 180 degrees relative to the start of the detection period Q2, an exposure is carried out to image the detection signal A180, followed by an exposure to image the detection signal B180. From the point in time when the phase is shifted by 270 degrees relative to the start of the detection period Q3, an exposure is carried out to image the detection signal A270, followed by an exposure to image the detection signal B270. The exposure time of detection signal B0 of detection period Q1 is compared here with the exposure time of detection signal A180 of detection period Q2, and the result shows that the exposure occurs at the same time. Accordingly, detection signal A180 of detection period Q2 can be substituted by detection signal B0 of detection period Q2. Likewise, detection signal B180 of detection period Q2 can be substituted by detection signal A0 of detection period Q0. Similarly, the exposure time of detection signal B90 in detection period Q1 can be compared with the exposure time of detection signal A270 in detection period Q3, and the result shows that the exposure occurs at the same time. Accordingly, detection signal A270 in detection period Q3 can be substituted by detection signal B90 in detection period Q1. Likewise, detection signal B270 in detection period Q3 can be substituted by detection signal A90 in detection period Q1. In this respect, the detection period Q0 and the detection period Q1 are defined as the detection period of a frame, and the detection signal A0 and the detection signal B0 are captured during the detection period Q0, as shown in Fig. 12. The detection signal A0, acquired during detection period Q0, can be used as detection signal B180. Similarly, the detection signal B0, acquired during detection period Q0, can be used as detection signal A180. Accordingly, in this case, during detection period Q0, detection signal A0, detection signal B0, detection signal A180, and detection signal B180 can be considered as if they had actually been acquired. Similarly, the detection signal A90, acquired during detection period Q1, can be used as detection signal B270. Furthermore, the detection signal B90, acquired during detection period Q1, can be used as detection signal A270. Accordingly, in this case, during detection period Q1, detection signal A90, detection signal B90, detection signal A270, and detection signal B270 can be considered as if they had actually been acquired. Therefore, the 2-tap 2-phase case described with reference to Fig. 12 can also be treated like the 2-tap 4-phase case described with reference to Fig. 10. The value I and the value Q in expression (1) in the 2-tap 4-phase described with reference to Fig. 10 are expressed as in expression (4) shown below. [Mathematical formula 4] The value I is obtained by subtracting the detection signal B0 from the detection signal A0, and the value Q is obtained by subtracting the detection signal B90 from the detection signal A90. Since the value I and the value Q are determined or acquired, the phase difference θ can be calculated according to expression (1) and the distance D can be calculated according to expression (2) as in the 2-tap system case described above. <1-Tap 4-Phase> Now, a case is briefly described in which the distance to a predetermined object is measured with a tap 51 (the configuration of a pixel 50 with a tap 51 is not shown) using irradiated light with four phase differences or via reception of light with four phase differences (which may be referred to here as "1-tap 4-phase"). If the imaging sequence in the temporal direction is illustrated in the same way as in Fig. 10 in the case of 1-tap 4-phase, the sequence shown in Fig. 13 is obtained. During a detection period Q0, the value C0 is determined in the expression (1) described above. During a detection period Q1, the value C90 is determined in the expression (1) shown above. During a detection period Q2, the value C180 is determined in the expression (1) shown above. During a detection period Q3, the value C270 is determined in the expression (1) shown above. In the case of the 1-tap system, the value I and the value Q are expressed in expression (1) above as in expression (5) below. [Mathematical formula 5] Since the value I and the value Q are determined, the phase difference θ can be calculated according to expression (1) as in the case of the 2-tap system described above, and the distance D can be calculated according to expression (2). The present technology can be applied to cases 2-type 4-phase, 2-tap 2-phase and 1-tap 4-phase, as described above. <Fliegende Pixel> A false detection occurring near the edge of a measured object within its environment is now described. A pixel that is falsely detected near the edge of an object can be referred to as a faulty pixel, flying pixel, or similar. As shown in Figures 14 and 15, this paper discusses a case in which there are two objects in a three-dimensional environment and the positions of the two objects are measured using the distance measuring device 10. Figure 14 is a diagram showing the positional relationship between a foreground object 101 and a background object 102 in an xz-plane. Figure 15 is a diagram showing the positional relationship between the foreground object 101 and the background object 102 in an xy-plane. The xz-plane shown in Fig. 14 is the plane when the foreground object 101, the background object 102 and the distance measuring device 10 are viewed from above, and the xy-plane shown in Fig. 15 is a plane that lies in a direction perpendicular to the xz-plane, and is the plane when the foreground object 101 and the background object 102 are viewed from the distance measuring device 10. Referring to Fig. 14, when the distance measuring device 10 is used as the reference, the foreground object 101 is located on the side closer to the distance measuring device 10, and the background object 102 is located on the side farther away from the distance measuring device 10. The foreground object 101 and the background object 102 are also within the field of view of the distance measuring device 10. The field of view of the distance measuring device 10 is indicated by a dashed line 111 and a dashed line 112 in Fig. 14. One side of the foreground object 101, or the side on the right in Fig. 14, is a border 103. There is a possibility that a flying pixel will appear near this border 103. Referring to Fig. 15, the distance measuring device 10 performs an imaging of the foreground object 101 and the background object 102 in a situation where these objects overlap. In such a case, there is a possibility that flying pixels may appear on the upper side (referred to as edge 104) and the lower side (referred to as edge 105) of the foreground object 101. In this case, a flying pixel is a pixel that is detected as belonging to an edge region of the foreground object 101, or as being detected at such a distance that the pixel belongs neither to the foreground object 101 nor to the background object 102. Fig. 16 is a diagram in which the foreground object 101 and the background object 102 are represented by the pixels corresponding to the image shown in Fig. 14. A pixel group 121 consists of the pixels detected by the foreground object 101, and a pixel group 122 consists of the pixels detected by the background object 102. Pixels 123 and 124 are flying pixels and are falsely detected pixels. Pixels 123 and 124 are located at the edges between foreground object 101 and background object 102, as shown in Fig. 16. Both of these floating pixels can belong to either foreground object 101 or background object 102. Alternatively, only one of these floating pixels can belong to foreground object 101, while the other belongs to background object 102. Pixels 123 and 124 are detected as floating pixels and processed accordingly, as shown, for example, in Fig. 17. Referring to Fig. 17, pixel 123 (Fig. 16) is modified to a pixel 123' belonging to pixel group 121, which is associated with foreground object 101, and pixel 124 (Fig. 16) is modified to a pixel 124' belonging to pixel group 122, which is associated with background object 102. The process for detecting faulty pixels, such as flying pixels, in this way is described in more detail below. <Erster Prozess in Bezug auf eine Detektion fliegender Pixel> Referring to Fig. 18, a first process for detecting flying pixels will now be described. Flying pixel detection is performed in the filter unit 16 (Fig. 1). Again referring to Fig. 1, a depth map and a confidence map of the filter unit 16 are provided by the signal processing unit 13. The filter unit 16 detects flying pixels from the depth map (a group of pixels). In step S11, the filter unit 16 defines a process target (evaluation target) pixel in the provided depth map. In step S12, a check is performed to determine whether the difference between the distance (depth) of the target pixel and the distance (depth) of an environment pixel is equal to or greater than a threshold value. Referring to Fig. 16, for example, a floating pixel such as pixel 123 is located away from pixel group 123 and pixel group 122. In other words, pixel 123 is located at a distance from the pixels surrounding it. Therefore, in a case where the distance between pixel 123 and an environment pixel (such as a pixel in pixel group 121) is calculated, the difference is assumed to be a predetermined value or greater. The determination process in step S12 is described further below. An example of a case where a directional derivative around a point corresponding to the process target pixel is used to detect a flying pixel is described below. Note that directional derivatives in several directions can be considered in the description below. For the sake of simplicity, however, the vertical and horizontal directions are used as examples. It is understood, however, that the same principle can be applied to directions other than the vertical and horizontal. Although one case in which directional derivatives are used is described here as an example, other methods can also be adopted within the present technology. Here, "P" represents the pixel being evaluated in the depth map, and "a" represents the selected direction in the plane. In this case, da(P) is the value of the derivative in direction "a" at pixel "P". If the absolute value |da(P)| of the directional derivative and |da + π(P)| are greater than predefined thresholds in direction "a", and the signs of da(P) and |da + π(P)} are the same, the pixel is detected as a floating pixel. Note that (a + π) indicates a direction opposite to a. Using the directional derivative, it is also possible to detect a pixel with noise. If the pixel currently being evaluated has a depth value that differs significantly from the depth values ​​of all pixels adjacent to the pixel currently being evaluated, or if at least one directional derivative is greater than a predefined threshold and at least two directional derivatives have opposite signs, the pixel is detected as a pixel with noise. Detection of a pixel with noise is also described herein. However, if the first process is performed with respect to the detection of flying pixels, the detection of a pixel with noise, described below, is a process that can be omitted if necessary. For each pixel, a detection of a pixel with noise and / or a detection of a flying pixel can be performed in any desired number of directions. These directions preferably cover a unit circle, which is a circle with a radius equivalent to one pixel. Typically, where i = 1 to n, a set of directions {a_i} can be used, and the following expression is satisfied: A directional derivative can be estimated from a finite difference. In Fig. 19, pixel 150 is pixel “P”, which is evaluated using pixels 152, 154, 156, and 158, corresponding to pixels “T” above pixel “P”, “L” to the left, “R” to the right, and “B” below pixel “P”. The values ​​of pixels 152, 154, 156 and 158 can be used to determine whether pixel 150 is a floating pixel or not, and whether the pixel is a pixel with noise or not, in two directions, the directions at angles of 0° and π / 2 (the horizontal and vertical directions). In the case of horizontal and vertical directions, the detection of a pixel with noise is performed according to the following expressions: Sign(R - P) ≠ Sign(P - L) or Sign(T - P) ≠ Sign(P - B). These expressions are defined as expression (6). In expression (6), Th represents the threshold. In expression (6), P represents the depth value of pixel 150, T represents the depth value of pixel 152, L represents the depth value of pixel 154, R represents the depth value of pixel 156, and B represents the depth value of pixel 158. Note that the same applies to expression (7) shown later. Furthermore, in the description below, each value is described as either less than or greater than the threshold. However, for example, each value could be equal to or greater than the threshold, or equal to or less than the threshold. The expression (|R - P| < Th and |L - P| < Th) is the expression to determine whether the absolute value of the depth value difference between pixel 150 and pixel 156, which lies on the right side of pixel 150 (Fig. 19), is less than the threshold Th and the absolute value of the depth value difference between pixel 150 and pixel 154, which lies on the left side of pixel 150, is less than the threshold Th. If a determination is made according to this expression, a check is also performed to determine whether Sign(R - P) ≠ Sign(P - L) holds true. The expression Sign(R - P) ≠ Sign(P - L) is used to determine whether the positive / negative sign of the value obtained by subtracting the depth value of pixel 150 from the depth value of pixel 156 is equal to the positive / negative sign of the value obtained by subtracting pixel 150 from pixel 154. If both of these conditions are met, the target pixel is detected as a pixel with noise. If processing is performed with attention to the vertical direction, (|T - P| < Th and |B - P| < Th) and sign(T - P) ≠ sign(P - B) are used. The expression (|T - P| < Th and |B - P| < Th) is used to determine whether the absolute value of the depth difference between pixel 150 and pixel 152, which lies above pixel 150 (Fig. 19), is less than the threshold Th, and whether the absolute value of the depth difference between pixel 150 and pixel 158, which lies below pixel 150, is less than the threshold Th. If a determination is made according to this expression, a check is also performed to determine whether Sign(T - P) ≠ Sign(P - B) holds true. The expression Sign(T - P) ≠ Sign(P - B) is used to determine whether the positive / negative sign of the value obtained by subtracting the depth value of pixel 150 from the depth value of pixel 152 is not equal to the positive / negative sign of the value obtained by subtracting the depth value of pixel 150 from the depth value of pixel 158. If both of these conditions are met, the target pixel is detected as a pixel with noise. The detection of a flying pixel is performed according to the following expressions: and sign(R - P) = sign(P - L) or sign(T - P) = sign(P - B). These expressions are defined as expression (7). In expression (7), Th represents the threshold and k represents a predetermined weighting factor. The expression (|R - P| > kTh and |L - P| > kTh) is the expression for determining whether the absolute value of the depth value difference between pixel 150 and pixel 156, which lies on the right side of pixel 150 (Fig. 19), is greater than the value obtained by multiplying the threshold Th by the weighting factor k, and whether the absolute value of the depth value difference between pixel 150 and pixel 154, which lies on the left side of pixel 150, is greater than the value obtained by multiplying the threshold Th by the weighting factor k. If a determination is made according to this expression, a check is also performed to determine if Sign(R - P) = Sign(P - L) holds true. The expression Sign(R - P) = Sign(P - L) is used to determine whether the positive / negative sign of the value obtained by subtracting the depth value of pixel 150 from the depth value of pixel 156 is equal to the positive / negative sign of the value obtained by subtracting the depth value of pixel 150 from the depth value of pixel 154. If both of these conditions are met, the target pixel is detected as a flying pixel. If processing is performed with attention to the vertical direction, (|T - P| > kTh and |B - P| > kTh) and sign(T - P) = sign(P - B) are used. The expression (|T - P| > kTh and |B - P| > kTh) is used to determine whether the absolute value of the depth difference between pixel 150 and pixel 152, which lies above pixel 150 (Fig. 19), is greater than the value obtained by multiplying the threshold Th by the weighting factor k, and whether the absolute value of the depth difference between pixel 150 and pixel 158, which lies below pixel 150, is greater than the value obtained by multiplying the threshold Th by the weighting factor k. If a determination is made according to this expression, a check is also performed to determine if Sign(T - P) = Sign(P - B) holds true. The expression Sign(T - P) = Sign(P - B) is used to determine whether the positive / negative sign of the value obtained by subtracting the depth value of pixel 150 from the depth value of pixel 152 is the same as the positive / negative sign of the value obtained by subtracting the depth value of pixel 150 from the depth value of pixel 158. If both of these expressions are true, the target pixel is detected as a flying pixel. Note that as an alternative to the above detection of a flying pixel or a pixel with noise, instead of expression (6) and expression (7), expression (8) shown below can be used. The expression (|L - R| > Th and |T - B| > Th) determines whether the absolute value of the depth difference between pixel 156, located to the right of the target pixel 150, and pixel 154, located to the left, is greater than the threshold Th, and whether the absolute value of the depth difference between pixel 152, located above the target pixel 150, and pixel 158, located below it, is greater than the threshold Th. If this expression is true, the target pixel is detected as a floating pixel. According to expression (8), the value between the two pixels surrounding the pixel being evaluated is used. Therefore, the threshold Th can be greater than the threshold Th in expressions (6) and (7) shown above. In step S12 (Fig. 18), the filter unit 16 determines whether the target pixel is a floating pixel or not, based on expression (7). In step S12, a determination process is performed to ascertain whether the depth value (distance) difference from an environment pixel is greater than (equal to or greater than) the threshold value. Although a floating pixel can only be detected by this process in step S12, there is a possibility that a pixel that is not a floating pixel may be detected as one. For example, it is possible that a plane located near a direction parallel to the direction vector of the distance measuring device 10, and at a large distance, might be detected as a flying pixel. The detection of a flying pixel from such a plane in the process in step S12 will lead to a false detection. The determination process in step S13 is performed as a mechanism to correct such a false detection carried out in step S12. If it is determined that the distance difference between the process target pixel and the surrounding pixel is equal to or greater than (or greater than) the threshold in step S12, in other words, if it is determined that there is a possibility of a flying pixel, the process proceeds to step S13. In step S13, a check is performed to determine whether the confidence or reflectance difference between the target pixel and the surrounding pixel is equal to or greater than the threshold value. Confidence is a value calculated according to expression (3). Reflectance is a value calculated according to expression (9) shown below. Note that, as described above, confidence could, for example, be the sum of the absolute values ​​of I and Q in an expression other than expression (3). The reflectance is the value obtained by multiplying the confidence by the square of the distance (depth value). In step S13, either confidence or reflectance can be used. Alternatively, a mechanism can be provided in which a specific index is introduced, and the use of confidence and reflectance can be switched between. If the determination process in step S13 is performed using the confidence level, the determination process is carried out based on expression (10) shown below. The confidence level is calculated based on expression (3) shown above. In expression (10) Pc represents the confidence of pixel 150, Tc represents the confidence of pixel 152, Lc represents the confidence of pixel 154, Rc represents the confidence of pixel 156 and Bc represents the confidence of pixel 158. The expression (|Rc - Pc| > Thc and |Lc - Pc| > Thc) determines whether the absolute value of the difference between the confidence of pixel 150 and the confidence of pixel 156, located to the right of pixel 150 (Fig. 19), is greater than a threshold value Thc, and whether the absolute value of the difference between the confidence of pixel 150 and the confidence of pixel 154, located to the left of pixel 150, is greater than the threshold value Thc. If this expression is satisfied, the target pixel is detected (confirmed) as a flying pixel. If processing is performed with attention to the vertical direction, (|Tc - Pc| > Th, and |Bc - Pc| > Thc) is used. The expression (|Tc - Pc| > Thc and |Bc - Pc| > Thc) determines whether the absolute value of the difference between the confidence of pixel 150 and the confidence of pixel 152, which lies above pixel 150 (Fig. 19), is greater than the threshold Thc, and whether the absolute value of the difference between the confidence of pixel 150 and the confidence of pixel 158, which lies below pixel 150, is greater than the threshold Thc. If this expression is satisfied, the target pixel is detected (confirmed) as a flying pixel. If the determination process in step S13 is carried out using the reflectance, the determination process is performed based on expression (11) shown below. The reflectance is calculated based on expression (9) shown above. In expression (11) Pr represents the reflectance of pixel 150, Tr represents the reflectance of pixel 152, Lr represents the reflectance of pixel 154, Rr represents the reflectance of pixel 156 and Br represents the reflectance of pixel 158. The expression (|Rr - Pr| > Thr and |Lr - Pr| > Thr) determines whether the absolute value of the difference between the reflectance of pixel 150 and the reflectance of pixel 156, located to the right of pixel 150 (Fig. 19), is greater than a threshold value Thr, and whether the absolute value of the difference between the reflectance of pixel 150 and the reflectance of pixel 154, located to the left of pixel 150, is greater than the threshold value Thr. If this expression is satisfied, the target pixel is detected (confirmed) as a flying pixel. In a case where processing is performed with attention to the vertical direction, (|Tr - Pr| > Thr and |Br - Pr| > Thr) is used. The expression (|Tr - Pr| > Thr and |Br - Pr| > Thr) determines whether the absolute value of the difference between the reflectance of pixel 150 and the reflectance of pixel 152, which lies above pixel 150 (Fig. 19), is greater than the threshold value Thr, and whether the absolute value of the difference between the reflectance of pixel 150 and the reflectance of pixel 158, which lies below pixel 150, is greater than the threshold value Thr. If this expression is satisfied, the target pixel is detected (acknowledged) as a flying pixel. The determination process in step S13 can be either a determination using confidence or a determination using reflectance. If the confidence is calculated by the signal processing unit 13 (Fig. 1) and provided to the filter unit 16, the filter unit 16 performs the determination process using the provided confidence. Alternatively, if the filter unit 16 performs the determination process using reflectance, the reflectance is calculated using the confidence provided by the signal processing unit 13, and the determination process is performed using the calculated reflectance. The threshold Thc in expression (10) and the threshold Thr in expression (11) are fixed values ​​and are values ​​that have been predetermined. Note that, although the explanation above includes a check to determine whether a value is greater than the threshold Thc (threshold Thr), a check can also be performed to determine whether a value is equal to or greater than the threshold Thc (threshold Thr). In step S13, a check is performed to determine whether the confidence or reflectance difference between the process target pixel and an environment pixel is greater than the threshold or not, and if it is determined that the difference is greater than the threshold, the process proceeds to step S14. In step S14, the filter unit 16 confirms that the process target (evaluation target) pixel is a floating pixel. The pixel confirmed as floating is then changed to a pixel belonging to pixel group 121 of foreground object 101 or pixel group 122 of background object 102, or it is deleted, as described above with reference to Fig. 17. Thus, the pixel is corrected. If, on the other hand, step S13 determines that the confidence or reflectance difference between the target process pixel and the environment pixel is equal to or less than the threshold value, the process proceeds to step S15. The process proceeds to step S15 if, in step S12, it is determined that the target pixel is not a floating pixel, if, in step S13, it is determined that the target pixel is not a floating pixel even though it was determined that there is a possibility of a floating pixel, or if, in step S14, it is confirmed that the target pixel is a floating pixel. In step S15, a check is performed to determine whether there are any unprocessed pixels. If step S15 determines that there are unprocessed pixels, the process returns to step S11 and the processes are repeated thereafter. If step S15 determines that there are no unprocessed pixels, the first process regarding the detection of flying pixels comes to an end. In this way, a pixel that could be a floating pixel is detected in step S12, and a verification is performed in step S13 to confirm that the pixel is indeed a floating pixel. That is, a floating pixel is detected (confirmed) in two stages. This reduces false detections of floating pixels and allows for higher accuracy in their detection. <Zweiter Prozess in Bezug auf eine Detektion fliegender Pixel> The first process for detecting flying pixels was described using an exemplary case where the threshold Thc (expression 10) and the threshold Thr (expression 11) are fixed values. However, a second process for detecting flying pixels is now described using an exemplary case where the threshold Thc (expression 10) and the threshold Thr (expression 11) are variable values. Fig. 20 is a flowchart to explain the second process in relation to the detection of flying pixels. The processes in steps S31 and S32 are similar to the processes in steps S11 and S12 (Fig. 18). A target pixel is defined, and a check is performed to determine whether the distance difference between the pixel and an environment pixel exceeds a threshold value. This detects a pixel that might be a flying pixel. In step S33, the threshold is calculated. Procedures for setting the threshold are described later. This threshold is either the threshold Thc (expression 10) or the threshold Thr (expression 11). After the threshold is set in step S33, the process is executed in step S34 without the set threshold. In step S34, a check is performed to determine whether the confidence or reflectance difference between the target pixel and an environment pixel is greater than the threshold. The process in step S34 is similar to the process in step S13 (Fig. 18), but differs in that the threshold calculated in step S33 is used as the threshold Thc (expression 10) or the threshold Thr (expression 11). The second process for detecting flying pixels is similar to the first, except that a threshold is set adaptively, and a determination to confirm a flying pixel is performed using this threshold. Steps S34 to S36 are executed in a manner similar to the processes in steps S13 to S15 (Fig. 18). The procedure for calculating the threshold in step S33 is further explained here. First method for setting the threshold The threshold is set at a value that is x times the average of the confidence values ​​or reflectances of the top, bottom, left, and right pixels. For example, x can be set to a value such as 0.1. Referring to Fig. 19, in a case where the determination process in step S34 is performed based on confidences, the average confidence value of pixel 152 above pixel 150 (the process target), pixel 158 below the process target, pixel 154 to the left of the process target, and pixel 156 to the right of the process target are calculated. Then, x times the average confidence value is set as the threshold Thc. Referring to Fig. 19, in a case where the determination process in step S34 is carried out based on reflectances, the average reflectance of pixel 152 above pixel 150 (the process target), pixel 158 below the process target, pixel 154 to the left of the process target, and pixel 156 to the right of the process target are calculated. Then, x times the average reflectance is set as the threshold Thr. Note that in a case where processing is performed using pixels oriented in one of the vertical and horizontal directions, as in expression (10) and expression (11), the threshold can be set using pixels oriented in that one direction. That is, although the top, bottom, left, and right pixels are used in the example above, the threshold can be calculated using pixels oriented vertically or pixels oriented horizontally. Note that the same applies to the determination methods described below. Although the top, bottom, left, and right pixels are used in the example cases described below, the threshold can, of course, be calculated using the pixels above and below the process target or the pixels to the right and left of the process target. Second method for setting the threshold The threshold is set at a value that is x times the standard deviation of the confidences or reflectances of the top, bottom, left, and right pixels. For example, x can be set to a value such as 0.2. Referring to Fig. 19, in a case where the determination process in step S34 is performed based on confidences, the standard deviation of the confidence of pixel 152 above pixel 150 (the process target), the confidence of pixel 158 below the process target, the confidence of pixel 154 to the left of the process target, and the confidence of pixel 156 to the right of the process target are calculated. Then, x times the standard deviation of the confidences is set as the threshold Thc. Referring to Fig. 19, in a case where the determination process in step S34 is performed based on reflectances, the standard deviation of the reflectance of pixel 152 above pixel 150 (the process target), the reflectance of pixel 158 below the process target, the reflectance of pixel 154 to the left of the process target, and the reflectance of pixel 156 to the right of the process target are calculated. Then, x times the standard deviation of the reflectances is set as the threshold Thr. Third procedure for setting the threshold The threshold is set to a value that is x times the difference between the largest and smallest values ​​of the confidences or reflectances of the top, bottom, left, and right pixels. For example, x could be set to a value such as 0.2. Referring to Fig. 19, in a case where the determination process in step S34 is performed based on confidence levels, the largest and smallest values ​​below the confidence level of pixel 152 above pixel 150 (the process target), pixel 158 below the process target, pixel 154 to the left of the process target, and pixel 156 to the right of the process target are detected. The difference between the largest and smallest values ​​is then calculated. Furthermore, x times this difference is set as the threshold Thc. Referring to Fig. 19, in a case where the determination process in step S34 is based on reflectances, the highest and lowest values ​​below the reflectance of pixel 152 above pixel 150 (the process target), the reflectance of pixel 158 below the process target, the reflectance of pixel 154 to the left of the process target, and the reflectance of pixel 156 to the right of the process target are detected. The difference between the highest and lowest values ​​is then calculated. Furthermore, x times this difference is set as the threshold value Thr. Fourth method for setting the threshold The threshold is set at a value that is x times the confidence level or reflectance of the target pixel. For example, x can be set at a value such as 0.1. Referring to Fig. 19, in a case where the determination process in step S34 is carried out using the confidence, an x-fold confidence of pixel 150 is set as the process target as the threshold Thc. Referring to Fig. 19, in a case where the determination process in step S34 is carried out using the reflectance, an x-fold of the reflectance of pixel 150 is set as the process target as the threshold Thr. Fifth method for setting the threshold The sum of the squares of the distances between two pixels, the target pixel and an adjacent pixel, is set as the threshold. The fifth method for setting the threshold is implemented in a case where the determination process in step S34 is performed using the reflectance. If the threshold is set using pixel 150 as the process target (Fig. 19) and pixel 152 (Fig. 19) which is above pixel 150, the sum of the square (d1502) of the distance (d150) of pixel 150 and the square (d1522) of the distance (d152) of pixel 152 is set as the threshold Thr. Note that although an exemplary case has been described herein in which the pixel located above the process target pixel is used, it is of course possible to use a pixel that is located at a different adjacent position than the position above the process target pixel. The derivation process according to the fifth determination procedure is now briefly described. In the description below, d represents a distance (low value), c represents a true confidence value, c' represents a measured confidence value, r represents a true reflectance value, and r' represents a measured reflectance value. An error in confidence and an error in reflectance can be expressed as in expression (13) shown below. [Mathematical formula 6] For example, the relationship between the error, the measured value and the true value of pixel 150 can be expressed as the process goal, and the relationship between the error and the measured value of the neighboring pixel 152 can be expressed as shown in Fig. 21. The upper part of Fig. 21 is a relationship diagram of pixel 150 (process target). If the measured value r150' of the reflectance of pixel 150 is placed in the middle, the true value r150 falls between (r150' - d1502) and (r150' + d1502). The lower part of Fig. 21 is a relationship diagram of pixel 152 (adjacent pixel). If the measured value r152' of the reflectance of pixel 152 is placed in the middle, the true value r152 falls between (r152' - d1522) and (r152' + d1522). It is assumed that the true value r of the reflectance of pixel 150 falls within a region A where the relationship diagram of pixel 150 (process target) and the relationship diagram of pixel 152 (adjacent pixel) overlap. Region A exists if expression (14) shown below is satisfied. That is, area A exists if the absolute value of the difference between the measured value r150' of the reflectance of pixel 150 and the measured value r152' of the reflectance of pixel 152 is less than the sum of the square of the distance d150 of pixel 150 and the square of the distance d152 of pixel 152. If expression (14) is satisfied, it can be assumed that the true value r exists. This means that if expression (14) is not satisfied, there is a possibility that the true value r does not exist. A time when expression (14) is not satisfied is a time when, for example, expression (15) is satisfied. If expression (15) is satisfied, it is assumed that there is a possibility that the true value r does not exist and the target pixel could be a faulty pixel. In this case, it is assumed that the pixel could be a floating pixel. Expression (15) is now compared with expression (11). Here again is a part of expression (11) (the expression in the case where the vertically oriented pixels are used). The focus is, for example, on (|Tr-Pr| > Thr) in expression (11). Tr represents a measured value of the reflectance of pixel 152 (Fig. 19) and corresponds to the measured value r152'. Pr represents a measured value of the reflectance of pixel 150 (Fig. 19) and corresponds to the measured value r150'. |Tr - Pr| is therefore |r150' - r152'|. That is, the left side of expression (11) can be considered equal to the left side of expression (15). Accordingly, the threshold Thr on the right-hand side of expression (11) can be (d1502+ d1522), which is the right-hand side of expression (15). As described above, according to the fifth method for setting the threshold, the sum of the squares of the distances between two pixels that are the process target pixel and an adjacent pixel can be set as the threshold Thr. Although the first five determination methods described above are examples of procedures for setting the threshold, these are only examples and do not imply any limitations. Therefore, the threshold can be set using any other method. Since a defective pixel is detected and confirmed through two determination processes as described above, it is possible to reduce false detection of defective pixels. This further improves the accuracy of distance measurements. <Beispielhafte Konfiguration einer elektronischen Einrichtung> The distance measuring device 10 described above can, for example, be mounted in an electronic device such as a smartphone, a tablet device, a mobile phone, a personal computer, a game console, a television receiver, a portable device, a digital still camera or a digital video camera. Fig. 22 is a block diagram showing an exemplary configuration of a smartphone as an electronic device in which the distance measuring device 10 is mounted as a distance measuring module. As shown in Fig. 22, a smartphone 201 contains a distance measuring module 202, an imaging device 203, a display 204, a loudspeaker 205, a microphone 206, a communication module 207, a sensor unit 208, a touch panel 209, and a control unit 210, which are interconnected via a bus 211. Furthermore, a CPU in the control unit 210 executes a program to implement functions as an application processing unit 221 and an operating system processing unit 222. The distance measuring device 10 in Fig. 1 is used for the distance measuring module 202. For example, the distance measuring module 202 is arranged in the front surface of the smartphone 201 and performs a distance measurement for the user of the smartphone 201 in order to output the depth value of the surface shape of the user's face, hand, finger or the like as a measurement result. The imaging device 203 is located on the front surface of the smartphone 201 and captures an image of the user by imaging the smartphone 201 as the object. Note that, although not shown in the drawing, the imaging device 203 can also be located on the rear surface of the smartphone 201. The display 204 shows a control screen for performing processing with the application processing unit 221 and the operating system processing unit 222, an image captured by the imaging device 203, or the like. The loudspeaker 205 and the microphone 206 output the voice from the other end and capture or record the user's voice when, for example, a voice call is made with the smartphone 201. The communication module 207 conducts communication via a communication network. The sensor unit 208 detects speed, acceleration, proximity, and the like, and the touch panel 209 detects operation by touch, which is carried out by the user on an operating screen displayed on the display 204. The application processing unit 221 performs processing to provide various services through the smartphone 201. For example, the application processing unit 221 can perform a process to generate a face using computer graphics that virtually reproduces the user's expression and display the face on the display 204, based on the depth provided by the distance measurement module 202. The application processing unit 221 can also perform a process to generate three-dimensional shape data of a three-dimensional object, based on the depth provided by the distance measurement module 202. The operating system processing unit 222 performs a process to implement the basic functions and operations of the smartphone 201. For example, the operating system processing unit 222 can perform a process to authenticate a user's face based on the depth value provided by the proximity sensor module 202 and unlock the smartphone 201. Furthermore, the operating system processing unit 222 performs a process to recognize a user gesture based on the depth value provided by the proximity sensor module 202 and then, for example, performs a process to input various operations according to the gesture. <aufzeichnungsmedium> The series of processes described above can be performed using hardware or software. If the series of processes is to be performed using software, the program that constitutes the software is installed on a computer. This computer could be, for example, a computer integrated into specialized hardware, or it could be a general-purpose personal computer or similar device capable of performing various functions with different types of installed programs. Figure 23 is a block diagram illustrating an exemplary hardware configuration of a computer that performs the series of processes described above according to a program. In the computer, a central processing unit (CPU) 501, a read-only memory (ROM) 502, and a random-access memory (RAM) 503 are interconnected by a bus 504. An input / output interface 505 is also connected to the bus 504. An input unit 506, an output unit 507, a memory unit 508, a communication unit 509, and a drive 510 are connected to the input / output interface 505. The input unit 506 consists of a keyboard, a mouse, a microphone, and the like. The output unit 507 consists of a display, a speaker, and the like. The storage unit 508 consists of a hard disk, non-volatile memory, or the like. The communication unit 509 consists of a network interface or the like. The drive 510 drives a removable recording medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory. In a computer with the configuration described above, for example, a CPU 501 loads a program stored in the memory unit 508 into the RAM 503 via the input / output interface 505 and the bus 504 and executes the program, so that the series of processes described above is carried out. The program to be executed by the computer (the CPU 501) can, for example, be recorded on the removable recording medium 511 as a packed medium and the like, and then made available. Alternatively, the program can be made available via a wired or wireless transmission medium such as a local network, the Internet, or digital satellite transmission. The program can be installed in the computer via the input / output interface 505 into the memory unit 508 when the removable recording medium 511 is inserted in the drive 510. Alternatively, the program can be received from the communication unit 509 via a wired or wireless transmission medium and installed in the memory unit 508. In other cases, the program can be installed beforehand in the ROM 502 or the memory unit 508. Note that the program to be executed by the computer may be a program to perform processes in chronological order according to the sequence described in this description, or a program to perform processes in parallel, or to perform a process as needed, such as when called. <Maschinelles Lernen nutzende beispielhafte Konfiguration> In a configuration for which the technology according to the present disclosure (the present technology) is used, a machine learning-utilizing configuration can be adopted. For example, a machine learning-utilizing configuration can be adopted in an electrical device equipped with the distance measuring device 10 described above. Fig. 24 is a diagram showing an exemplary configuration of an unclaimed example of an electronic device for which the present technology is used. An electronic device 601 is a device such as a smartphone, a tablet, a mobile phone, a personal computer, a game console, a television receiver, a portable device, a digital still camera, or a digital video camera. For example, the electronic device 601 can be the smartphone 201 shown in Fig. 22. The electronic device 601 includes a sensor 602, a controller 603, and an output unit 604. The sensor 602 includes a distance measuring unit 611 and a processing unit 612. The distance measuring unit 611 is designed to correspond to the distance measuring device 10 shown in Fig. 1 and includes a lens 11, a light-receiving unit 12, a signal processing unit 13, a light-emitting unit 14, and a light emission control unit 15. The distance measuring unit 611 illuminates an object with light and receives the light reflected from the object to measure the distance to the object. The distance measuring unit 611 acquires distance measurement information obtained by measuring the distance to the object and provides this distance measurement information to the processing unit 612. For example, the distance measurement information can be a depth map containing data (distance information) related to a distance measurement for each pixel. In the depth map, a group of pixels can be represented by an XYZ coordinate system (such as a Cartesian coordinate system) or a polar coordinate system. In some cases, the depth map contains data related to a correction target pixel. Three-dimensional data can also be generated using the depth map. For example, the three-dimensional data might be expressed as a point cloud, a series (a data structure) of XYZ coordinate points, a mesh (surface), or the like. Note that the distance measurement information can include a luminance value, confidence information, and the like, in addition to depth information such as a distance value (depth value). Correction target pixels include a faulty pixel, a floating pixel, and the like. As described above with reference to Figures 14, 15, 16 to 17, a floating pixel is a pixel that belongs neither to the foreground object 101 nor to the background object 102, located at the boundary between the foreground object 101 and the background object 102. A faulty pixel is described below as an example of a correction target pixel. The processing unit 612 is, for example, a processor such as a CPU. The processing unit 612 has the functions corresponding to the filter unit 16. That is, in the distance measuring device 10 in Fig. 1, the filter unit 16 identifies and corrects a defective pixel. However, some or all of the corresponding processes can be performed by the processing unit 612 as described below. Note that instead of the distance measuring unit 611, the processing unit 612 can have the functions corresponding to the signal processing unit 13. As described above, in the sensor 602, the distance measuring unit 611 has a function for receiving the reflected light, which is the irradiated light reflected from the object, and outputting a signal obtained from the reception. The processing unit 612 has a function for performing signal processing using the signal output by the distance measuring unit 611. Although further details will be described later with reference to Fig. 40, the processing unit 612 can be located within the sensor 602 or can be located in the electronic device 601 or a server. Furthermore, the processing unit 612 can perform all the processes within the sensor 602 or it can perform processes in a distributed manner, with some processes being performed by a processor in the sensor and the remaining processes then being performed by the processor of the electronic device 601 or a server. The processing unit 612 performs a process using a machine-learned learning model 621 on at least part of the distance measurement information provided by the distance measurement unit 611 and outputs a distance measurement information after the faulty pixel contained in the distance measurement information has been corrected (this distance measurement information may be referred to herein as corrected distance measurement information). The process utilizing the 621 learning model comprises a first process to correct a faulty pixel using the depth map containing the faulty pixel as input, and a second process to output the corrected depth map. The first process comprises an identification process to identify a defective pixel and a correction process to correct the defective pixel identified through the identification process. That is, in the first process, the identification process is performed as the first step, and the correction process is performed as the second step. The identification or correction process may involve a process utilizing the 621 learning model. In the following, where applicable, a process utilizing machine learning, such as a process using the 621 learning model, will be referred to as the machine learning process. For example, in the identification process, the learning model 621 is a learning model that receives an input of a depth map containing a faulty pixel and outputs position information about the faulty pixel contained in the depth map. Similarly, in the correction process, the learning model 621 is a learning model that receives an input of the depth map containing the faulty pixel and position information about the faulty pixel, and outputs a corrected depth map or position information about the corrected faulty pixel. Note that methods for generating a learning model are described in detail later with reference to Figures 29 to 31, 36, 37, and others. Furthermore, the identification or correction process is not limited to machine learning; a process similar to the one described above with reference to Fig. 18 or Fig. 20 can be used. That is, in the identification process, instead of machine learning, a process can be performed to identify the faulty pixel based on the result of a distance information threshold determination with respect to the target pixel in the depth map, or a confidence information or reflectance information threshold determination. Furthermore, in the correction process, signal processing can be performed instead of machine learning to change the target pixel in the depth map to the correct position or to delete the target pixel. Thus, the faulty pixel can be corrected. In the following, we refer to signal processing such as that described above with reference to Fig.The process described in 18 or Fig. 20 may also be referred to as the regular processing. Note that processing unit 612 can generate three-dimensional data based on the corrected depth map. In this three-dimensional data generation process, machine learning can be used, or signal processing (regular processing) can be used to generate three-dimensional data. That is, machine learning can be used in at least one of the identification process, the correction process, and the three-dimensional data generation process. The corrected distance measurement information obtained in this way is output by the processing unit 612 of the sensor 602 to the controller 603. The controller 603 is, for example, a processor such as a CPU. The processor of the controller 603 can be the same as the processor used in the processing unit 612 or different from it. The output unit 604 contains, for example, a display 622. The controller 603 executes software, such as an application and middleware, to cause the display 622 to show different types of information. For example, the controller 603 causes the display 622 to show different types of information based on the corrected distance measurement information output by the sensor 602. Note that the output unit 604 contains a communication module and can transmit different types of information from the controller 603 to other devices via a network. Next, with reference to Figs. 25, 26, 27 to 28, the processes to be carried out by the sensor 602 of the electronic device 601 in Fig. 24 are described. First, with reference to Fig. 25, the processing of distance measurement information is described. In step S111, the distance measuring unit 611 measures the distance to the object and records distance measurement information obtained through the measurement. In step S112, the processing unit 612 performs a learning model application process using the distance measurement information acquired by the distance measurement unit 611. In this learning model application process, a process utilizing the learning model 621 is performed on at least a portion of the distance measurement information (for example, a depth map still requiring correction), and the distance measurement information (for example, the corrected depth map) is acquired after the faulty pixel contained in the distance measurement information has been corrected. The learning model application process is described in detail later with reference to Fig. 26. In step S113, the processing unit 612 outputs the corrected distance measurement information obtained in the process of using a learning model to the controller 603. Next, with reference to Fig. 26, the process of using a learning model, corresponding to step S112 in Fig. 25, is described. In step S131, the processing unit 612 performs a process to correct faulty pixels using the distance measurement information. In this process, the faulty pixel is corrected, with the depth map containing the faulty pixel serving as input. The process for correcting defective pixels is described in detail here with reference to Fig. 27. In step S151, the processing unit 612 identifies the defective pixel contained in the input depth map. This identification process involves either machine learning or regular processing. If the machine learning process is performed as the identification process, the processing unit 612 uses a learning model 621 that receives an input of the depth map containing the defective pixel and outputs positional information about the defective pixel contained in the depth map. While the processing unit 612 inputs the depth map containing the defective pixel into the learning model 621, positional information about the defective pixel contained in the input depth map is output. Thus, the defective pixel can be identified. A method for generating a learning model to be used in the identification process is described in detail later with reference to Figures 29, 30 to 31. If a neural network is used as the machine learning tool in the machine learning process employed in the identification process, an autocoder is also used to identify the defective pixel contained in the input depth map. A method for generating the learning model in a case where an autocoder is used is described in detail later with reference to Figures 32 and 33. If regular processing is performed as the identification process, the processing unit 612 identifies the faulty pixel based on the result of a distance information threshold determination in relation to the process target pixel in the depth map or the result of a confidence information or reflectance information threshold determination. Specifically, as described above with reference to Fig. 18 or Fig. 20, a distance information (depth value) threshold determination is performed in regular processing to determine whether the distance difference between the target pixel (for example, pixel 123) in the depth map and an environment pixel (for example, a pixel in pixel group 121 around pixel 123) is greater than a predetermined threshold (step S12 in Fig. 18). If it is determined that the distance difference is greater than the predetermined threshold, a check is performed to determine whether the confidence or reflectance difference between the target pixel (for example, pixel 150) and an environment pixel (for example, pixel 156 to the right of pixel 150) is greater than a predetermined threshold (step S13 in Fig. 18).If the confidence or reflectance difference is determined to be greater than the predetermined threshold, the target process pixel is then identified as a defective pixel (step S14 in Fig. 18). After the defective pixel contained in the depth map is identified in step S151, the process continues to step S152. In step S152, the processing unit 612 corrects the identified defective pixel. This correction process involves either machine learning or regular processing. If the machine learning process is performed as the correction process, the processing unit 612 uses a learning model 621 that receives input of the depth map containing the defective pixel and the position information about the defective pixel, and outputs a corrected depth map or position information about the corrected defective pixel. While the processing unit 612 inputs the depth map containing the defective pixel and the position information about the defective pixel into the learning model 621, the corrected depth map or the position information about the corrected defective pixel is output. Thus, the defective pixel can be corrected. A method for generating the learning model to be used in the correction process is described in detail later with reference to Figures 36 and 37. If regular processing is performed as the correction process, processing unit 612 corrects the faulty pixel by changing the target pixel in the depth map to the correct position or by deleting the target pixel through signal processing. Specifically, as described above with reference to Fig. 18 or Fig. 20, in regular processing the target pixel identified (confirmed) as the faulty pixel is changed to a pixel belonging to pixel group 121 of foreground object 101 or pixel group 122 of background object 102, or it is deleted. Thus, the faulty pixel is corrected. As described above, in the process of correcting faulty pixels, the identification process to identify a faulty pixel contained in a depth map is performed as the first step, and the correction process to correct the identified faulty pixel is performed as the second step. Note that in a case where the machine learning process is used in both the identification and correction processes, these processes can be performed integrally (simultaneously). For example, if a neural network is used for machine learning, it is possible to correct a faulty pixel contained in an input depth map using an autocoder. A method for generating the learning model in a case where an autocoder is used is described in detail later with reference to Figures 34 and 35. After the identified faulty pixel is corrected in step S152, the process proceeds to step S132 in Fig. 26. In step S132, the processing unit 612 outputs a depth map (corrected depth map) that includes the faulty pixel corrected in the process of using a learning model. As described above, in the processing of distance measurement information, a process utilizing the learning model 621 is performed by the processing unit 612 on at least a portion of the distance measurement information acquired by the distance measurement unit 611. For example, the machine learning process can be used for at least either the identification process or the correction process, which are contained within the first process. If the machine learning process is not performed in either the identification or correction process at this time, regular processing can be carried out. Note that when processing distance measurement information, three-dimensional data can be generated using a corrected depth map. Fig. 28 illustrates the processing of distance measurement information in a case where three-dimensional data is to be generated. In steps S171 and S172 in Fig. 28, a process of using a learning model that utilizes the distance measurement information acquired by the distance measuring unit 611 is carried out, as in steps S111 and S112 in Fig. 25. In step S173, the processing unit 612 generates three-dimensional data based on the corrected depth map. This three-dimensional data generation process can utilize either machine learning or regular processing. For example, the three-dimensional data can be generated as a point cloud, a data structure, or something similar. In step S174, the processing unit 612 outputs the generated three-dimensional data to the controller 603. Note that the processing of distance measurement information described with reference to Fig. 28 can be carried out in a continuous machine learning process, i.e., end-to-end. Specifically, in the sensor 602 of the electronic device 601, the processing unit 612 can output three-dimensional data using the learning model 621 when distance measurement information (a depth map containing a defective pixel) is input from the distance measurement unit 611. The learning model 621 can be a generation model of a generative adversarial network (GAN), a variational autoencoder (VAE), or the like. As described above, in continuous machine learning, a learning process is performed by replacing the machine learning process, which requires multi-stage processes from receiving input data to outputting a result, with a neural network containing a multitude of layers that perform various processes. Using continuous machine learning, the processing unit 612, which receives the distance measurement information from the distance measurement unit 611, performs the machine learning process with the learning model 621 and outputs three-dimensional data. As described above, in the electronic device 601, the processing unit 612 of the sensor 602 performs a process that includes the identification process to identify a defective pixel contained in a depth map and the correction process to correct the identified defective pixel, as well as the processing of distance measurement information using the distance measurement information provided by the distance measurement unit 611. Furthermore, the processing of distance measurement information may include a three-dimensional data generation process to generate three-dimensional data. When processing distance measurement information, machine learning can be used in at least one of the processes: identification, correction, and three-dimensional data generation. This means that machine learning can be performed in all of these processes, or it can be performed in some of them while regular processing takes place in the remaining processes. Specifically, after machine learning or regular processing is performed in the identification process, it is then performed in the correction process, and finally in the three-dimensional data generation process. For the sake of simplicity, the identification process, the correction process, and the three-dimensional data generation process are represented by the letters "A", "B", and "C", respectively. The machine learning process and regular processing are represented by the numbers "1" and "2", respectively, and the letters and numbers are connected by a hyphen ("-"). In this case, a series of processes is carried out when processing distance measurement information as one of the combinations A-1 or A-2, B-1 or B-2, and C-1 or C-2. The description below explains in detail combinations A-1 (a case where the machine learning process is carried out during the identification process) and B-1 (a case where the machine learning process is carried out during the correction process). (A-1) Exemplary case in which the machine learning process is carried out in the identification process Referring to Figures 29, 30 to 31, a case is described in which the machine learning process is carried out during the identification process. If the machine learning process is carried out during the identification process, the learning model to be used in the machine learning process is generated beforehand by a learning device. Fig. 29 shows an exemplary configuration of an unclaimed example of a learning device for which the present technology is used. A learning device 701 generates a learning model using training data. The learning device 701 can be the same device as the electronic device 601 (such as a smartphone) or it can be a different device (such as a server). The learning device 701 includes a generation unit 703 and an output unit 704. The generation unit 703, for example, is a processor such as a CPU. The processor of the generation unit 703 can be the same as the processor of the processing unit 612 or the processor of the controller 603, or it can be different from them. Alternatively, all the processors of the generation unit 703, the processing unit 612, and the controller 603 can be the same processors. The generation unit 703 acquires or acquires training data. For example, the training data includes a depth map containing a defective pixel and a depth map identifying the position of the defective pixel. In addition to measurement data obtained through actual measurement, data generated by a simulator can be used as training data. For example, the simulator uses information relating to the 611 distance measurement unit (such as the sensor type and lens distortion) as parameters and processes data virtually (performing a process to generate a depth map that intentionally includes a faulty pixel at a specific location, or the like). As a result, a dataset is generated to identify a faulty pixel, such as a depth map containing a faulty pixel and a depth map identifying the position of the faulty pixel. The generation unit 703 creates a learning model using the acquired training data and provides the learning model to the output unit 704. For example, the generation unit 703 creates a learning model that receives input of a depth map containing a defective pixel and outputs positional information about the defective pixel in the depth map. The generation unit 703 can also perform a relearning process (an additional learning process) using data such as additional training data to update the learned learning model. In the description below, the creation and updating of the learning model are explained separately. However, it can also be said that a learning model is created by updating a learning model, and therefore the meaning of the expression "creating a learning model" includes "updating a learning model." The output unit 704 outputs the learning model provided by the generation unit 703. For example, the output unit 704 is a communication module and can transmit the learning model to another device via a network. Fig. 30 shows an example of a learning model generated by the generation unit 703 of the learning device 701 in Fig. 29. Fig. 30 represents a machine-learned learning model that uses a neural network and comprises three layers: an input layer, an intermediate layer, and an output layer. The learning model in Fig. 30 is a learning model generated using training data containing a depth map containing faulty pixels whose positions are identified, and is a learning model that receives an input of distance measurement information 721 (a depth map containing faulty pixels as indicated by circles in the drawing) and outputs position information 722 about the faulty pixels (coordinate information about the faulty pixels contained in the input depth map). The processing unit 612 of the electronic device 601 in Fig. 24 can use the learning model in Fig. 30 as the learning model 621 when the machine learning process is performed during the identification process. That is, using the learning model shown in Fig. 30, the processing unit 612 performs an arithmetic operation on the depth map containing the defective pixels, which is input to the input layer, at the intermediate layer with the learned parameters, in order to determine the positions of the defective pixels. Position information about the defective pixels contained in the input depth map is output by the output layer. In this arithmetic operation at the intermediate layer, if the position of a defective pixel is identified, it is possible to capture the position information about the defective pixel by identifying the defective pixel based on a result of a distance information threshold determination with respect to the process target pixel in the depth map and a result of a confidence information or reflectance information threshold determination as in the process described above with reference to Fig. 18 or Fig. 19. Specifically, a check is performed using the distance information (depth value) threshold determination to determine whether the distance difference between the target pixel in the depth map and an environment pixel is greater than a predetermined threshold (step S12 in Fig. 18). If the distance difference is determined to be greater than the predetermined threshold, a check is performed to determine whether the confidence or reflectance difference between the target pixel and an environment pixel is greater than a predetermined threshold (step S13 in Fig. 18). If the confidence or reflectance difference is determined to be greater than the predetermined threshold, the target pixel is then identified as a defective pixel (step S14 in Fig. 18). The positional information for a faulty pixel, output by the output layer, can be information that identifies the position of the faulty pixels whose positions have been determined. For example, for each piece of information used to identify a faulty pixel, the positional information for a faulty pixel can express the position of the faulty pixel in the XYZ coordinate system. Alternatively, the positional information for a faulty pixel can be a map (an anomaly probability map) that represents the probability of a faulty pixel being present for each pixel. According to this anomaly probability map, for example, a pixel that is certain to belong to the foreground or background object will have a value close to 0, while a pixel with a high probability of being a faulty pixel will have a value close to 100. Referring to Fig. 31, a learning process for generating the learning model shown in Fig. 30 is described. In step S211, the generation unit 703 acquires training data. This training data includes a depth map containing defective pixels whose positions have been identified. In addition to measurement data, data generated by a simulator can also be used as training data. In step S212, using the acquired training data, the generation unit 703 creates a learning model that receives input of the depth map containing the defective pixels and outputs positional information about the defective pixels in the depth map. This learning model (Fig. 30) is used to identify the defective pixels in the depth map. In step S213, the output unit 704 outputs the learning model generated by the generation unit 703. This learning model (Fig. 30) is provided to the sensor 602 of the electronic device 601 and is used when the processing unit 612 performs the identification process to identify a defective pixel. As described above, in a case where the machine learning process is carried out in the identification process, a learning model is used that receives an input of a depth map containing faulty pixels and outputs position information about the faulty pixels contained in the depth map, so that the faulty pixels contained in the depth map can be identified. (A-1) Another exemplary case in which the machine learning process is carried out in the identification process. Referring to Figures 32 to 33, another exemplary case is described in which the machine learning process is carried out in the identification process. In this example, an autocoder is used in machine learning with a neural network to identify defective pixels. The autocoder contains an encoder and a decoder. One or more attribute values ​​contained in data are input into the autocoder. The encoder converts the input attribute set into a compressed expression. The decoder then decodes an attribute set from this compressed expression. At this point, the decoder generates data with a structure similar to that of the input data. During the learning process using the autocoder, normal data is input, and each parameter of the autocoder is optimized to minimize the difference between the input data and the reconfigured data (data generated by the autocoder). Therefore, after sufficient learning, the input data and the reconfigured data essentially have the same value. Figure 32 illustrates an exemplary method for identifying defective pixels using an autocoder. In Figure 32, the learning process is shown in the upper section, and the inference process is shown in the lower section. That is, the learning process is performed by the learning device 701 (Figure 29), and the inference process is performed by the electronic device 601 (Figure 24). The learning process and the inference process are included in the machine learning process and are performed by the same device if the learning device 701 and the electronic device 601 are configured as the same device. In the learning stage, the learning process is performed using an autocoder on a distance measurement information 741 input to the input layer, resulting in a distance measurement information 742 being output from the output layer. If a normal depth map containing no defective pixels is input as the distance measurement information 741, a depth map containing no defective pixels will be output as the distance measurement information 742. At this point, parameters optimized so that the input normal depth map and the reconfigured depth map have the same value can be extracted as feature sets 743. In the inference stage, the inference process is performed using a learning model (a neural network) that reflects the feature sets 743 obtained in the training stage, on a distance measurement information 751 input to the input layer, so that a distance measurement information 752 is output by the output layer. If a depth map containing defective pixels is input as the distance measurement information 751, the areas of the defective pixels cannot be reconfigured in the inference process, and a depth map that essentially contains no defective pixels is output as the distance measurement information 752. Accordingly, when the machine learning process is performed during the identification process, the processing unit 612 of the electronic device 601 in Fig. 24 can identify pixels with different positional information between the depth map input to the input layer and the depth map output by the output layer as defective pixels. For example, if each pixel in a depth map is expressed in an XYZ coordinate system, a pixel with a large difference in coordinates between the depth map containing defective pixels in the input layer and the depth map from which the defective pixels have disappeared in the output layer can be identified as a defective pixel. Note that in a case where a normal depth map is input as distance measurement information 751 into the input layer, a depth map containing no faulty pixels is output as distance measurement information 752 from the output layer. Accordingly, there are no pixels with a large difference in coordinates between the depth map at the input layer and the depth map at the output layer, and no faulty pixels are identified. Referring to Fig. 33, a defective pixel identification process using the learning model shown in Fig. 32 is described. The identification process shown in Fig. 33 corresponds to a case where the machine learning process is performed in step S151 in Fig. 27. In step S311, using a neural network (Fig. 32) that has learned a normal depth map with an autocoder, the processing unit 612 inputs a depth map containing faulty pixels as the learning model 621 into the input layer and outputs a depth map in which the faulty pixels have been corrected from the output layer. In step S312, the processing unit 612 identifies the defective pixels contained in the depth map based on the positional information in the depth map input to the input layer and the positional information in the depth map output by the output layer. For example, pixels with different positional information between the depth map input to the input layer and the depth map output by the output layer can be identified as defective pixels. As described above, in a case where a learning model utilizing an autocoder is used, when the machine learning process is performed in the identification process, the position information in the depth map input to the input layer and the position information in the depth map output by the output layer are used so that faulty pixels can be identified. (A-1, B-1) Exemplary case in which the machine learning process is carried out in a series of identification and correction processes. Referring to Figures 34 and 35, a case is described in which the machine learning process is carried out in the identification and correction processes, which are performed as a series of processes. In this example, an autocoder is also used in machine learning with a neural network to correct faulty pixels. Fig. 34 shows an exemplary method for correcting defective pixels using an autocoder. In Fig. 34, the learning process in the learning stage is shown in the upper part, and the inference process in the inference stage is shown in the lower part, as shown in Fig. 32. In the learning stage, if a normal depth map containing no defective pixels is input as distance measurement information 761 into the input layer, the learning process is performed, and a depth map containing no defective pixels is output by the output layer as distance measurement information 762. At this point, parameters optimized so that the input normal depth map and the reconfigured depth map have the same value can be extracted as feature sets 763. In the inference stage, if a depth map containing defective pixels is input as distance measurement information 771 into the input layer, the inference process is carried out using a training model (neural network) that reflects the feature sets 763 obtained in the training stage. However, the areas of the defective pixels cannot be reconfigured, and a depth map is obtained that essentially contains no defective pixels. In the inference process using the training model, the pixels with differing position information between the depth map input to the input layer and the depth map output by the output layer are corrected as defective pixels. For example, if each pixel in a depth map is expressed in an XYZ coordinate system, a pixel with a large difference in coordinates between the depth map containing erroneous pixels in the input stage and the depth map from which the erroneous pixels have disappeared in the output stage can be corrected as an erroneous pixel. In correcting erroneous pixels, as described above with reference to Figure 18 or Figure 20, the target pixel can be moved or offset relative to the foreground or background object, or deleted. This means that the processing unit 612 of the electronic device 601 in Fig. 24 performs the inference process shown in Fig. 34 to integrally (simultaneously) identify and correct the defective pixels contained in the depth map. While the inference process is being carried out in this way, a depth map in which the defective pixels have been corrected can be output as distance measurement information 772 from the output layer. Referring to Fig. 35, a process for correcting defective pixels using the learning model shown in Fig. 34 is described. The correction process shown in Fig. 35 corresponds to a case in which the machine learning process is carried out in step S131 in Fig. 26. In step S331, using a neural network (Fig. 34) that has learned a normal depth map with an autocoder, the processing unit 612 inputs a depth map containing faulty pixels as the learning model 621 into the input layer and outputs a depth map in which the faulty pixels have been corrected from the output layer. As described above, in a case where a learning model is used with an autocoder, if the machine learning process is performed in a series of identification and correction processes as a result of inputting a depth map containing faulty pixels into the input layer, a depth map in which the faulty pixels have been corrected is output from the output layer. Thus, the faulty pixels can be corrected. (B-1) Exemplary case in which a machine learning process is carried out in the correction process Referring to Figures 36 to 37, a case is described in which the machine learning process is carried out and used in the correction process. Fig. 36 shows an example of a learning model generated by the generating unit 703 of the learning device 701 in Fig. 29. In Fig. 36, the learning process in the learning stage is shown in the upper area and the inference process is shown in the lower area. In the learning stage, a depth map containing faulty pixels, position information about the faulty pixels, and a corrected depth map or position information about the corrected faulty pixels are used as training data. Figure 36 shows the learning process using these pieces of training data to create a learning model that receives an input of distance measurement information 781 (a depth map containing faulty pixels) and position information 782 about the faulty pixels (coordinate information about the faulty pixels) and outputs a distance measurement information 783 (a corrected depth map). Note that when correcting faulty pixels, as described above with reference to Fig. 18 or Fig. 20, the target pixel can be moved relative to the foreground or background object or deleted. While Fig. 36 illustrates a case where the output of the learning model is a corrected depth map, the learning process can also be performed where the output of the learning model is the positional information of the corrected faulty pixels. Regarding the positional information of the faulty pixels, coordinate information can be used that expresses the positions of the faulty pixels in an XYZ coordinate system. In the inference stage, the learning model generated during the learning process is used, and the inference process is performed on distance measurement information 791 (a depth map containing faulty pixels) and position information 792 about the faulty pixels (coordinate information about the faulty pixels), which are input into the input layer, so that a distance measurement information 793 (a corrected depth map) is output from the output layer. Note that instead of a corrected depth map, position information about the corrected faulty pixels can be output. Referring to Fig. 37, a learning process for generating the learning model shown in Fig. 36 is described. In step S411, the generation unit 703 acquires training data. This training data includes a depth map containing defective pixels and position information about the defective pixels, as well as a corrected depth map or position information about the corrected defective pixels. In addition to the measurement data, data generated by a simulator can also be used as training data. In step S412, using the acquired training data, the generation unit 703 creates a learning model that receives an input of a depth map containing faulty pixels and position information about the faulty pixels, and outputs a corrected depth map or position information about the corrected faulty pixels. This learning model (Fig. 36) is used to correct the faulty pixels contained in the depth map. In step S413, the output unit 704 outputs the learning model generated by the generation unit 703. This learning model (Fig. 36) is provided to the processing unit 612 and the like of the sensor 602 in the electronic device 601 and is used in the correction process to correct defective pixels. As described above, in a case where the machine learning process is performed in the correction process, a learning model is used that receives an input of a depth map containing faulty pixels and position information about the faulty pixels and outputs a corrected depth map or position information about the corrected faulty pixels, so that the identified faulty pixels can be corrected. Note that in another exemplary case, where the machine learning process is performed during the correction process, erroneous pixels can be corrected using a generative adversarial network (GAN). A GAN consists of two networks: a generation network (a generator) and an identification network (a discriminator). In the learning stage, a normal depth map is used, and training is performed to generate a depth map from random values. In the inference stage, when a depth map containing erroneous pixels is input, the generation network (the generator) creates a depth map using parameters designed to produce a depth map similar to the input depth map, and the generated depth map can be designated as the corrected depth map.Alternatively, the identification network (the discriminator) determines the loss in a depth map, allowing defective pixels to be detected and corrected. Note that, although a GAN was described as an example, a neural network such as a U-network can also be used. <Beispielhafte Nutzung einer Tiefenkarte> Note that the above description primarily explained cases where the machine learning process is used. However, after defective pixels contained in a depth map have been identified in regular processing, a depth map in which the identified defective pixels have been corrected can be generated, and the corrected depth map can then be used. Fig. 38 is a diagram showing another exemplary configuration of an embodiment of an electronic device for which the present technology is used. The electronic device 601 in Fig. 38 differs from the electronic device in Fig. 24 in that it contains a sensor 631 instead of the sensor 602. The sensor 631 in Fig. 38 differs from the sensor 602 in Fig. 24 in that it contains a processing unit 641 instead of the processing unit 612. Note that in the electronic device 601 in Fig. 38, the components corresponding to those of the electronic device 601 in Fig. 24 are designated with the same reference numerals as those in Fig. 24, and these are not explained here. The processing unit 641 generates a corrected depth map using distance measurement information provided by the distance measurement unit 611. Referring to Fig. 39, a process for generating a corrected depth map is now described. In step S511, the distance measurement unit 611 performs a process related to the detection of flying pixels. In this process, as described above with reference to Fig. 18 or Fig. 20, a process can be carried out to identify, based on the result of a distance information threshold determination with respect to the target pixel in the depth map, or a confidence information or reflectance information threshold determination. In step S512, processing unit 641 corrects the faulty pixel identified in the depth map and generates a corrected depth map. When correcting faulty pixels, the target pixel in the depth map is either moved to the correct position or deleted. In step S513, the processing unit 641 outputs the generated corrected depth map. The corrected depth map can, of course, be used within the electronic device 601 and can also be provided to an external device. For example, the uncorrected depth map, together with the corrected depth map, can be provided to the learning device 701. With this arrangement, the learning device 701 can generate a training model using the data set of the uncorrected depth map and the corrected depth map as the training data. <Prozessgegenstand im maschinellen Lernprozess> In the configuration utilizing machine learning described above, the processing unit 612 in the sensor 602 performs the machine learning process in the electronic device 601. However, the machine learning process is not necessarily performed by the processing unit 612 in the sensor 602, but can be performed by any other device. Fig. 40 shows an example of a device that performs the machine learning process in a case where the electronic device 601 is a mobile terminal. In Fig. 40, the electronic device 601 is a mobile terminal such as a smartphone and can be connected to another network, such as the Internet 653, via a core network 652 connected to a base station installed at a predetermined location. In Fig. 40, an edge server 651 for implementing mobile edge computing (MEC) is located closer to the mobile terminal, such as between the base station and the core network 652. Note that the edge server 651 can be located within the core network 652. Furthermore, a cloud server 654 is connected to the Internet 653. The edge server 651 and the cloud server 654 can perform various types of use-dependent processes. In Fig. 40, the machine learning process is performed by the electronic device 601, the sensor 602, the edge server 651, or the cloud server 654. In each device, the machine learning process is performed, for example, by a processor such as a program-executing CPU or by dedicated hardware. Note that, in addition to a CPU, a graphics processing unit (GPU), a general-purpose computing unit (GPGPU), a quantum computer, or the like can be used. One configuration in which the processing unit 612 in the sensor 602 performs the machine learning process is as described above. The external configuration of the sensor 602 can, for example, be a configuration as described in Fig. 41. In Fig. 41, the sensor 602 can be configured as a single-chip semiconductor device with a stacked structure in which a plurality of substrates are stacked. The sensor 602 is formed by stacking two substrates: a substrate 951 and a substrate 952. In Fig. 41, a distance measuring unit 911 is mounted on the upper substrate 951. Components from a distance measurement processing unit 912 to a distance measurement control unit 915 and components from a CPU 921 to an input interface 926 are mounted on the lower substrate 952. The distance measuring unit 911 and the components from the distance measurement processing unit 912 to the distance measurement control unit 915 form a distance measuring block 901. The components from the CPU 921 to the input interface 926 form a signal processing block 902. For example, the distance measuring block 901 corresponds to the distance measuring unit 611 in Fig. 24, and the signal processing block 902 corresponds to the processing unit 612 in Fig. 24. That is, in the stacked structure in which the upper and lower substrates are stacked, a chip for the machine learning process is mounted on the lower substrate 952. Since such a configuration is adopted, the distance measurement information acquired by the distance sensing block 901, which is designed for distance measurement and mounted on the upper substrate 951 or the lower substrate 952, is processed by the signal processing block 902, which is designed for the machine learning process and mounted on the lower substrate 952. Thus, a number of processes can be performed in the single-chip semiconductor device. Note that in a case where the signal processing block 902 corresponds to the processing unit 612 in Fig. 24, processes such as the machine learning process can be performed by the CPU 921. Since a number of processes are performed in this way within the single-chip semiconductor device, no information leaks out, which is preferable from a privacy standpoint. Furthermore, there is no need to transmit information. Thus, the signal processing block 902 can use the distance measurement information provided by the distance sensing block 901 to perform the machine learning process, such as the inference process, at high speed. For example, if the sensor is used in a game or similar application requiring real-time capabilities, sufficient real-time performance can be ensured. Furthermore, when the machine learning process is performed in the signal processing block 902, various types of metadata are passed from the controller 603 or a digital signal processor (DSP) in the electronic device 601.This allows processes to be reduced and electricity consumption to be lowered. Note that the sensor 602 can have a configuration that includes an additional or companion chip. For example, in a case where the sensor 602 comprises a first semiconductor element, which includes the distance measuring unit 611, containing the light-receiving unit 12, the light-emitting unit 14, and the like, and a second semiconductor element, which includes the processing unit 612 designed as a signal processing circuit, the second semiconductor element can be configured as a companion chip. The first and second semiconductor elements can be integrated as a single module. Furthermore, the first and second semiconductor elements can be configured on the same substrate. As described above, the companion chip in sensor 602 contains the processing unit 612, which is designed as a signal processing circuit. Thus, the companion chip can perform both regular processing and machine learning. In this arrangement, even in cases where high-load signal processing is required in addition to regular processing or machine learning, the companion chip, which is specialized for such signal processing, can perform the processing. Since the processing unit 612, designed as a signal processing circuit, performs the signal processing in hardware, it can be carried out at a higher speed and with lower power consumption than in cases where signal processing is performed in software. Referring back to Fig. 40, the electronic device 601 contains the controller 603, similar to a CPU; however, the controller 603 can perform the machine learning process. The machine learning process includes the learning process or the inference process. If the controller 603 of the electronic device 601 performs the inference process, the inference process can be carried out without requiring any time after the distance measurement information has been acquired by the distance measuring unit 611. Thus, high-speed processing can be performed. Accordingly, if the electronic device 601 is used in a game or any other application that requires real-time capabilities, the user can operate it without experiencing any sense of unease caused by a delay.Furthermore, if the controller 603 of the electronic device 601 performs the machine learning process, the process can be carried out at a lower cost than in a case where the cloud server 654 is used. The processor of the Edge Server 651 can perform the machine learning process. Since the Edge Server 651 is located physically closer to the Electronic Device 601 than the Cloud Server 654, it is possible to achieve lower process latency. The Edge Server 651 also offers higher processing capabilities, such as higher operating speed, compared to the Electronic Device 601 and the Sensor 602. Therefore, the Edge Server 651 can be designed for general-purpose use and can perform the machine learning process whenever it successfully receives data, regardless of variations in the specifications and performance of the Electronic Device 601 and the Sensor 602. If the Edge Server 651 performs the machine learning process, it is possible to reduce the processing load on the Electronic Device 601 and the Sensor 602. The processor of the Cloud Server 654 can perform the machine learning process. The Cloud Server 654 has high processing capabilities, such as high operating speed, and can therefore be designed for general-purpose applications. If the Cloud Server 654 performs the machine learning process, it can be carried out regardless of variations in the specifications and performance of the electronic device 601 and the sensor 602. If it is difficult for the processing unit 612 in the sensor 602 or the controller 603 in the electronic device 601 to perform a machine learning process under high load, the Cloud Server 654 can perform the machine learning process under high load and feed the processing result back to the processing unit 612 in the sensor 602 or the controller 603 in the electronic device 601. As described above, the processor of the electronic device 601, sensor 602, edge server 651, or cloud server 654 has at least one function (some or all functions) of the processing unit 612 and the generation unit 703 to perform the machine learning process. The processor in each device can also perform all of the processes. Alternatively, after some of the processes have been performed by the processor in one device, the remaining processes can be performed by the processor in another device. Note that the machine learning process can be performed by a device other than those shown in Fig. 40. For example, the machine learning process can be performed by any other electronic device to which the electronic device 601 can be connected via wireless communication or the like. Specifically, in a case where the electronic device 601 is a smartphone, examples of other electronic devices could include a smartphone, a tablet device, a mobile phone, a personal computer, a game console, a television receiver, a portable device, a digital still camera, and a digital video camera. If a sensor mounted on a mobile structure such as an automobile, a sensor intended for use in a remotely controlled medical device, or the like, is used in a configuration other than that shown in Fig. 40, the machine learning process, such as the inference process, can still be applied, but low latency is required in these environments. In such a case, it is desirable for the controller 603, which is provided in the electronic device 601 (including devices in vehicles), to perform the machine learning process instead of the cloud server 654, which performs the machine learning process over a network.Furthermore, in a case where there are no environments to connect to a network such as the Internet, or in a case of a device used in an environment where no high-speed connection is possible, it is desirable that the controller 603 provided in the electronic device 601 also performs the machine learning process. Note that the machine learning method used in the machine learning process can be a neural network, deep learning, or something similar. A neural network is a model that mimics the circuitry of human cranial nerves and contains three types of layers: an input layer, an intermediate layer (a hidden layer), and an output layer. Deep learning, meanwhile, is a model that utilizes a neural network with a multi-layered structure and can learn a complex pattern hidden in a large dataset by repeatedly learning characteristics in each layer. Furthermore, examples of problems addressed in machine learning include supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning, and any of these can be used. For example, in supervised learning, sets of features are learned based on labeled training data that is provided. As a result, it becomes possible to derive a label attached to unknown data. In unsupervised learning, on the other hand, a large amount of unlabeled training data is analyzed, and sets of features are extracted. Based on the extracted sets of features, clustering or similar analysis is then performed. Consequently, it becomes possible to perform trend analysis and future predictions based on a vast amount of unknown data. Furthermore, semi-supervised learning is a method that combines supervised and unsupervised learning. It is a technique for repeatedly performing a learning process, automatically computing feature sets by providing a vast amount of training data in unsupervised learning after the feature sets have been learned in supervised learning. Reinforcement learning, on the other hand, is a method for addressing a problem in determining an action that an agent or resource should perform in a given environment by observing the current state. <Beispielhafte Konfiguration eines Systems> Data such as a training model and a depth map can, of course, be used in a single device and can also be exchanged between and used in a multitude of devices. Fig. 42 shows an exemplary configuration of a system comprising devices for which the present technology is used. In Fig. 42, electronic devices 601-1 to 601-N (where N is an integer 1 or greater), a learning device 701, a server 672 providing a learning model, a server 673 providing a depth map, and an application server 674 are connected to a network 671 and can exchange data with each other. The network 671 includes a mobile phone network, the Internet, and the like. For example, the learning device 701, the server 672 providing a learning model, the server 673 providing a depth map, and the application server 674 are each provided as the cloud server 654 shown in Fig. 40. The learning device 701 can generate a learning model using training data and make the learning model available to the electronic devices 601-1 to 601-N via the network 671. In each electronic device 601, the processing unit 612 performs the machine learning process using the learning model provided by the learning device 701. The learning model is not necessarily provided directly by the learning device 701 to each electronic device 601, but can, for example, be provided to each electronic device 601 via the learning model-providing server 672, which can provide various types of learning models. The learning model-providing server 672 can provide the learning model to a device other than the electronic devices 601. The electronic devices 601-1 to 601-N can provide processed data, such as an uncorrected depth map and a corrected depth map, to other devices via the network 671. An uncorrected depth map is the data output by the distance measuring unit 611 and is the data that still needs to be processed by the processing unit 612. A corrected depth map is the data output by the distance measuring unit 611 and is the data processed by the processing unit 612. Alternatively, the data to be provided by the electronic devices 601-1 to 601-N can include metadata or the like, obtained by performing a recognition process or the like. For example, the electronic devices 601-1 through 601-N can provide an uncorrected depth map and a corrected depth map to the learning device 701 via the network 671. The learning device 701 can generate a training model using the uncorrected and corrected depth maps collected as training data from the respective electronic devices 601. Depth maps are not necessarily provided directly to the learning device 701 by each electronic device 601, but can, for example, be provided to the learning device 701 via the depth map server 673, which can provide various types of depth maps. The depth map server 673 can provide depth maps to a device other than the learning device 701. The learning device 701 can update the learning model by performing a relearning process on the already generated learning model, adding the depth maps collected by the respective electronic devices 601 to the training data. If the user performs a correction operation on the corrected depth map (for example, if the user enters correct information) in an electronic device 601, data (feedback data) regarding the correction process can also be used in the relearning process. For example, by transmitting the feedback data from the electronic device 601 to the learning device 701, the learning device 701 can perform the relearning process using the feedback data from the electronic device 601 and update the learning model.Note that an application provided by application server 674 can be used if the user performs a corrective operation. Although Fig. 42 herein presents an exemplary case in which the learning device 701 is the cloud server 654, the electronic devices 601 and the learning device 701 are not necessarily separate devices and may be the same device. If the electronic devices 601 and the learning device 701 are the same device, the controller 603 of each electronic device 601 or the processing unit 612 of the sensor 602, for example, has the functions of the generating unit 703 of the learning device 701. If the electronic devices 601 and the learning device 701 consist of the same device, a common processor for performing the respective functions may be included, or separate processors for the respective functions may be included. The learning process, or the relearning process, is the machine learning process, and, as described above with reference to Fig. 40, the learning or relearning process can be performed by the electronic device 601, the sensor 602, the edge server 651, or the cloud server 654. If the learning process is performed in the electronic device 601 or the sensor 602, information does not leak out, thus protecting privacy. Furthermore, if the electronic device 601 or sensor 602 updates the learning model by performing the relearning process using a depth map or feedback data, the learning model can be improved within the device. If the relearning process is performed, the electronic device 601 or the sensor 602 can transmit the updated learning model to the server 672 providing a learning model, so that the updated learning model is made available to the other electronic devices 601 and the other sensors 602. As a result, the updated learning model can be shared among the respective electronic devices 601 and the respective sensors 602.Alternatively, the electronic device 601 or the sensor 602 can transmit difference information about the newly learned learning model (information about the difference between the learning model before the update and the updated learning model) as update information to the learning device 701. The learning device 701 can then generate an improved learning model based on the updated information and make the improved learning model available to the other electronic devices 601 or the other sensors 602. Because such difference information is exchanged, privacy can be protected and communication costs can be reduced. If the learning or relearning process is performed on the cloud server 654, such as the learning device 701, the processes can be carried out regardless of variations in specifications and the performance of the electronic device 601 and the server 602. For example, the cloud server 654 can perform the learning process using depth maps provided as training data by the electronic device 601 or the depth map-providing sensor 673. Furthermore, when the electronic device 601 or the sensor 602 acquires a depth map or feedback data, the data is transmitted to the cloud server 654 so that the cloud server 654 can perform the relearning process. The learning model updated in the relearning process is provided to the other electronic devices 601 and the other sensors 602 by the learning model-providing server 672.As a result, the updated learning model can be shared among the respective electronic devices 601 and the respective sensors 602. The application server 674 is a server that can provide various types of applications over the network 671. These applications include, for example, an application that provides a function utilizing a learning model or a depth map. Any electronic device 601 can implement a function utilizing a learning model or a depth map by running the application downloaded from the application server 674 over the network 671. Note that each electronic device 601 can store a learning model previously provided by the learning device 701 into a storage unit such as a semiconductor memory, or can acquire a learning model from an external device such as the server 672 providing a learning model via the network 671 when the machine learning process such as the inference process is carried out. As described above, in a system comprising devices for which the present technology is used, data such as a training model and a depth map are exchanged and distributed among the respective devices, and various services can be provided using this data. For example, electronic devices can be manufactured and provided in which a training model generated by the training device 701 is installed. Distance measurement information acquired by the sensor 602 can also be input into the training model generated by the training device 701, and a depth map in which defective pixels have been corrected can be generated and provided as output.Furthermore, the generated learning model, the generated corrected depth map, and three-dimensional data generated based on the corrected depth map are stored on a readable recording medium, enabling the manufacture and provision of a storage medium that stores the learning model, the corrected depth maps, and the three-dimensional data, as well as electronic devices that each contain the storage medium. The storage medium can be non-volatile memory such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, or it can be volatile memory such as static random-access memory (SRAM) or dynamic random-access memory (DRAM). <Beispielhafte Anwendung für ein System für endoskopische Chirurgie > The technology (the present technology) according to the present disclosure (the present technology) can be used for various products. For example, the technology according to the present disclosure can be used for a system for endoscopic surgery. Fig. 43 is a diagram illustrating an exemplary configuration of a system for endoscopic surgery for which the technology (the present technology) according to the present disclosure can be used. Fig. 43 illustrates a situation in which a surgeon (a medical professional) 11131 is performing a surgical procedure on a patient 11132 on a patient bed 11133 using an endoscopic surgery system 11000. As shown in the drawing, the endoscopic surgery system 11000 comprises an endoscope 11100, other surgical instruments 11110 such as a pneumoperitoneum tube 11111 and an energy treatment instrument 11112, a support arm device 11120 that carries the endoscope 11100, and a trolley 11200 on which various types of devices for endoscopic surgery are mounted. The endoscope 11100 comprises a lens tube 11101, which extends a predetermined length from the head end and is to be inserted into a body cavity of the patient 11132, and a camera head 11102, which is connected to the base end of the lens tube 11101. In the example shown in the drawing, the endoscope 11100 is designed as a so-called rigid or inflexible endoscope with the rigid or inflexible lens tube 11101. However, the endoscope 11100 can also be designed as a so-called soft or flexible endoscope, which includes a flexible lens tube. An opening is provided at the head end of the lens tube 11101, into which an objective lens is inserted. A light source device 11203 is connected to the endoscope 11100, and the light generated by the light source device 11203 is guided to the head end of the lens tube 11101 through a light guide extending inside the lens tube 11101 and is emitted via the objective lens towards a current observation target in the patient's body cavity 11132. It is particularly noteworthy that the endoscope 11100 can be a straight-viewing endoscope, an oblique-viewing endoscope, or a lateral-viewing endoscope. Within the camera head 11102, an optical system and an imaging element are provided, and reflected light (observation light) from the observation target is focused by the optical system onto the imaging element. The observation light is photoelectrically converted by the imaging element, and an electrical signal corresponding to the observation light or an image signal corresponding to the observation image is generated. The image signal is transmitted as raw data (ROH) to a camera control unit (CCU) 11201. The CCU 11201 consists of a central processing unit (CPU), a graphics processing unit (GPU), or the like, and collectively controls the operation of the endoscope 11100 and the display device 11202. Furthermore, the CCU 11201 receives an image signal from the camera head 11102 and subjects the image signal to various types of image processing, such as a demosaicing process, to display an image based on the image signal. Under the control of the CCU 11201, the display device 11202 shows an image based on the image signal subjected to image processing by the CCU 11201. The light source device 11203 consists of a light source such as a light-emitting diode (LED) and supplies the endoscope 11100 with irradiation light to image the surgical site. An input device 11204 is an input interface for the 11000 system for endoscopic surgery. The user can use the input device 11204 to input various types of information and instructions into the 11000 system for endoscopic surgery. For example, the user can enter an instruction or similar to change imaging conditions (such as the type of irradiation light, magnification, and focal length) for the 11100 endoscope. A device 11205 for controlling a treatment instrument controls the energy treatment instrument 11112 for tissue cauterization, incision, closure of a blood vessel, or the like. To ensure the field of view of the endoscope 11100 and the working space for the surgeon, a pneumoperitoneum device 11206 introduces gas into a body cavity of the patient 11132 via the pneumoperitoneum tube 11111 to expand the body cavity. A recording device 11207 is a device that can record various types of information about the surgical procedure. A printer 11208 is a device that can print various types of information relating to the surgical procedure in various formats, such as text, images, graphics, and the like. It is particularly noteworthy that the light source device 11203, which provides irradiation light to the endoscope 11100 to image the surgical site, can consist, for example, of an LED, a laser light source, or a white light source that is a combination of an LED and a laser light source. If a white light source consists of a combination of RGB laser light sources, the output intensity and timing of each color (each wavelength) can be controlled with high precision. Accordingly, the white balance of an image acquired by the light source device 11203 can be adjusted. Alternatively, in this case, laser light from each of the RGB laser light sources can be emitted onto the observation target in a time-division multiplexing manner, and the imaging element of the camera head 11102 can be controlled synchronously with the irradiation time.Thus, images can be captured in time-division multiplexing according to their respective RGB colors. According to this method, a color image can be obtained without using any color filter for the imaging element. Furthermore, the control of the light source device 11203 can be configured to change the intensity of the emitted light at predetermined time intervals. The control of the imaging element of the camera head 11102 is synchronized with the change in light intensity, and images are captured in time-division multiplexing and then combined. This allows for the generation of a high dynamic range image without black areas or white spots. Furthermore, the light source device 11203 can also be designed to provide light of a predetermined wavelength band compatible with special-light observation. In special-light observation, for example, light with a narrower band than the irradiation light (or white light) used in normal observation is emitted, exploiting the wavelength dependence of light absorption in body tissue. As a result, a so-called narrowband observation (narrowband imaging) is performed to image a predetermined tissue, such as a blood vessel in a mucosal surface layer or the like, with high contrast. Alternatively, special-light observation can be used to perform fluorescence observation, obtaining an image by means of fluorescence generated by irradiation with excitation light.In fluorescence observation, excitation light is emitted to a body tissue so that the fluorescence of the body tissue can be observed (autofluorescence observation). Alternatively, a reagent such as indocyanine green (ICG) is injected locally into body tissue, and excitation light corresponding to the fluorescence wavelength of the reagent is emitted to the body tissue so that a fluorescence image can be obtained. The light source device 11203 can be designed to provide narrowband light and / or excitation light compatible with such observation using special light. Fig. 44 is a block diagram illustrating examples of the functional configurations of the camera head 11102 and the CCU 11201 shown in Fig. 43. The camera head 11102 contains a lens unit 11401, an imaging unit 11402, a control unit 11403, a communication unit 11404, and a camera head control unit 11405. The CCU 11201 contains a communication unit 11411, an image processing unit 11412, and a controller 11413. The camera head 11102 and the CCU 11201 are connected to each other by a transmission cable 11400. The lens unit 11401 is an optical system provided at a connection point with the lens tube 11101. Observation light captured at the head end of the lens tube 11101 is guided to the camera head 11102 and enters the lens unit 11401. The lens unit 11401 consists of a combination of multiple lenses, including a zoom lens and a focusing lens. The imaging unit 11402 can consist of a single imaging element (a so-called single-plate type) or of multiple imaging elements (a so-called multi-plate type). If the imaging unit 11402 is of the multi-plate type, for example, image signals corresponding to the respective RGB colors can be generated by the individual imaging elements and combined to obtain a color image. Alternatively, the imaging unit 11402 can be designed to contain a pair of imaging elements to capture image signals for the right and left eyes, compatible with a three-dimensional (3D) display. Because a 3D display is performed, the surgeon 11131 can more accurately determine the depth of body tissue at the surgical site.It is particularly worth mentioning that if the imaging unit 11402 is of the multi-plate type, a large number of lens units 11401 are provided for the respective imaging elements. Furthermore, the imaging unit 11402 is not necessarily located in the camera head 11102. For example, the imaging unit 11402 can be located directly behind the objective lens in the lens tube 11101. The control unit 11403 consists of an actuator and, under the control of the camera head control unit 11405, moves the zoom lens and the focus lens of the lens unit 11401 by a predetermined distance along an optical axis. This arrangement allows the magnification and the focal point of the image captured by the imaging unit 11402 to be appropriately adjusted. The communication unit 11404 consists of a communication device for transmitting and receiving various types of information to and from the CCU 11201. The communication unit 11404 transmits the image signal received as raw data from the imaging unit 11402 to the CCU 11201 via the transmission cable 11400. The communication unit 11404 also receives a control signal from the CCU 11201 to control the camera head 11102 and provides the control signal to the camera head control unit 11405. The control signal contains information about imaging conditions, such as information to determine the frame rate of captured images, information to determine the exposure value at the time of imaging, and / or information to determine the magnification and focal point of captured images. It is particularly worth mentioning that the above imaging conditions, such as frame rate, exposure value, magnification, and focus, can be appropriately specified by the user or automatically set by the CCU 11201 controller 11413 based on a captured image signal. In the latter case, the 11100 endoscope features an automatic exposure (AE) function, an autofocus (AF) function, and an automatic white balance (AWB) function. The camera head control unit 11405 controls the operation of the camera head 11102 on the basis of a control signal that is received via the communication unit 11404 from the CCU 11201. The communication unit 11411 consists of a communication device for transmitting and receiving various types of information to and from the camera head 11102. The communication unit 11411 receives an image signal transmitted from the camera head 11102 via the transmission cable 11400. Furthermore, the communication unit 11411 also transmits a control signal to the camera head 11102 for controlling its operation. The image signal and the control signal can be transmitted by means of electrical communication, optical communication, or the like. The image processing unit 11412 performs various types of image processing on an image signal, which is raw data transmitted by the camera head 11102. The controller 11413 performs various types of control regarding the display of an image of the surgical area or the like, captured by the endoscope 11100, and an image obtained by imaging the surgical site or the like. For example, the controller 11413 generates a control signal to control the camera head 11102. Furthermore, the controller 11413 also causes the display device 11202 to display a captured image, depicting a surgical site or the like, based on the image signal that has undergone image processing by the image processing unit 11412. The controller 11413 can recognize the respective objects depicted in the captured image using various image recognition techniques. Since the distance measuring device 10 or the electronic device 601 is used for the imaging unit 11402 and the image processing unit 11412, the controller 11413 can, for example, more accurately recognize a surgical instrument such as forceps, a specific body part, bleeding, mist during the use of the energy treatment instrument 11112, and the like, by more precisely detecting the shape and color of the edge of each object contained in the captured image.When the display device 11202 is instructed to display the captured image, the controller 11413 can cause the display device 11202 to overlay various types of information on the image of the surgical site to support a surgical intervention, using the recognition result. Since the information is overlaid and displayed to support a surgical intervention, and thus presented to the surgeon 11131, this reduces the workload for the surgeon 11131 and allows the surgeon 11131 to proceed safely with the surgical intervention. The transmission cable 11400, which connects the camera head 11102 and the CCU 11201, is an electrical signal cable compatible with electrical signal communication, an optical fiber compatible with optical communication, or a composite cable thereof. In the example shown in the drawing, communication is carried out using wired transmission cable 11400. However, communication between the camera head 11102 and the CCU 11201 can be carried out wirelessly. <Beispielhafte Anwendungen für bewegliche Strukturen> The technology (the present technology) according to the present disclosure can be used for various products. For example, the technology according to the present disclosure can be implemented as a device that is mounted in any type of movable structure, such as an automobile, an electric vehicle, a hybrid electric vehicle, a motorcycle, a bicycle, a personal mobility device, an aircraft, a drone, a ship, or a robot. Fig. 45 is a block diagram schematically representing an exemplary configuration of a vehicle control system, which is an example of a system for controlling a moving structure for which the technology according to the present disclosure can be used. A vehicle control system 12000 comprises a plurality of electronic control units connected via a communication network 12001. In the example shown in Fig. 45, the vehicle control system 12000 includes a powertrain control unit 12010, a body control unit 12020, an external information detection unit 12030, an internal vehicle information detection unit 12040, and an overall control unit 12050. Furthermore, the functional components of the overall control unit 12050 are illustrated as a microcomputer 12051, an audio / video output unit 12052, and an interface (I / F) 12053 of the network in the vehicle. The 12010 drive system control unit controls the operation of devices related to a vehicle's drive system according to various programs. For example, the 12010 drive system control unit acts as a control device for components such as a drive force generation device (e.g., an internal combustion engine or a drive motor) to generate the vehicle's drive force, a drive force transmission mechanism to transfer drive force to the wheels, a steering mechanism to adjust the vehicle's steering angle, and a braking device to generate the vehicle's braking force. The 12020 body control unit controls the operation of various devices mounted on a vehicle body according to different programs. For example, the 12020 body control unit can function as a keyless entry system, a smart key system, an automatic window regulator, or a control unit for various lights such as headlights, taillights, brake lights, turn signals, fog lights, and the like. In this case, the 12020 body control unit can receive radio waves from a portable device that replaces a key or signals from various switches.The body control unit 12020 receives inputs of these radio waves or signals and controls the door locking device, the automatic window lifting device, the lights and the like of a vehicle. The external information detection unit 12030 detects information about the external environment of the vehicle, which is equipped with the vehicle control system 12000. For example, an imaging unit 12031 is connected to the external information detection unit 12030. The external information detection unit 12030 causes the imaging unit 12031 to take an image of the vehicle's external environment and receives the captured image. Based on the received image, the external information detection unit 12030 can perform an object detection process to detect a person, vehicle, obstacle, traffic sign, marking on the road surface, or the like, or it can perform a distance detection process. The 12031 imaging unit is an optical sensor that receives light and outputs an electrical signal corresponding to the amount of light received. The 12031 imaging unit can output this electrical signal as an image or as distance measurement information. Furthermore, the light received by the 12031 imaging unit can be visible light or invisible light, such as infrared radiation. The distance measuring device 10 or the electronic device 601 is used for the imaging unit 12031. Using the distance measuring device 10 or the electronic device 601, it is possible to identify a pixel (a faulty pixel or a floating pixel) that appears to exist between the foreground and the background and to perform a correction process by repositioning the pixel to a correct position in the foreground or background, or by deleting the pixel. Accordingly, object detection can be performed correctly, and false object detection can be reduced. As a result, it is possible, for example, to prevent an accident during automated driving. A machine learning process can be used for object detection. Specifically, a bounding box or a semantic segmentation technique can be used for object detection.If a machine-learned learning model is used in a process to correct the target pixel (a faulty pixel or a floating pixel), the inference process using the learning model is performed within the sensor, enabling accurate object detection with minimal latency. This can significantly contribute to preventing accidents. The 12040 vehicle interior information detection unit detects information about the interior of the vehicle. For example, a 12041 driver condition detector, which detects the driver's state, is connected to the 12040 vehicle interior information detection unit. The 12041 driver condition detector includes, for example, a camera that captures an image of the driver, and based on the information detected by the 12041 driver condition detector, the 12040 vehicle interior information detection unit can calculate the driver's fatigue level or concentration level, or determine whether the driver is dozing off. Based on external / internal information acquired by the external information detection unit 12030 or the vehicle interior information detection unit 12040, the microcomputer 12051 can calculate the control target value of the drive force generation device, steering mechanism, or braking device and issue a control command to the drive system control unit 12010. For example, the microcomputer 12051 can perform cooperative control to implement the functions of an Advanced Driver Assistance System (ADAS), which include vehicle collision avoidance or impact mitigation, vehicle following based on distance between vehicles, constant vehicle speed driving, vehicle collision warning, vehicle lane departure warning, and the like. Furthermore, the microcomputer 12051 can also perform cooperative control to carry out automated driving or the like, to drive autonomously without being dependent on the intervention of the driver, by controlling the drive force generating device, the steering mechanism, the braking device or the like on the basis of information about the vehicle's environments, the information being acquired by the unit 12030 for detecting external information or the unit 12040 for detecting information from inside the vehicle. The microcomputer 12051 can also issue a control command to the body system control unit 12020 based on external information acquired by the external information detection unit 12030. For example, the microcomputer 12051 controls the headlight according to the position of the vehicle ahead or the oncoming vehicle detected by the external information detection unit 12030, and performs cooperative control to achieve a glare-free effect by switching from high beam to low beam or the like. The audio / video output unit 12052 transmits an audio output signal and / or a video output signal to an output device that can convey information visually or audibly to the vehicle's occupant(s) or the vehicle's external environment. In the example shown in Fig. 45, the output devices are a loudspeaker 12061, a display unit 12062, and an instrument panel 12063. The display unit 12062 can, for example, include an on-board display and / or a head-up display. Fig. 46 is a diagram illustrating an example of installation positions of the imaging units 12031. In Fig. 46, imaging units 12101, 12102, 12103, 12104 and 12105 are included as the imaging unit 12031. Imaging units 12101, 12102, 12103, 12104, and 12105 are, for example, located at the following positions: the front of a vehicle 12100, a side mirror, the rear bumper, a tailgate, the upper portion of a front windshield inside the vehicle, and the like. Imaging unit 12101, located at the front, and imaging unit 12105, located at the upper portion of the front windshield inside the vehicle, primarily capture images in front of the vehicle 12100. Imaging units 12102 and 12103, located at the side mirrors, primarily capture images of the sides of the vehicle 12100. Imaging unit 12104, located at the rear bumper or a tailgate, primarily captures images behind the vehicle 12100.The imaging unit 12105, located at the top of the front windscreen inside the vehicle, is primarily used to detect a vehicle, pedestrian, obstacle, traffic light, traffic sign, lane or the like in front of the vehicle. Note that Fig. 46 shows an example of the acquisition and imaging areas of the imaging units 12101 to 12104. An imaging area 12111 indicates the imaging area of ​​the imaging unit 12101 located at the front of the vehicle, imaging areas 12112 and 12113 indicate the imaging areas of the imaging units 12102 and 12103 located at the respective side mirrors, and an imaging area 12114 indicates the imaging area of ​​the imaging unit 12104 located at the rear bumper or a tailgate. For example, image data acquired by the imaging units 12101 to 12104 are superimposed to obtain a bird's-eye view image of the vehicle 12100 as seen from above. At least one of the imaging units 12101 to 12104 can have a function for acquiring distance information. For example, at least one of the imaging units 12101 to 12104 can be a stereo camera containing a variety of imaging elements, or it can be an imaging element containing pixels for detecting phase differences. For example, based on the distance information obtained from the imaging units 12101 to 12104, the microcomputer 12051 calculates the distances to the respective three-dimensional objects within the imaging areas 12111 to 12114 and temporal changes in the distances (relative speeds with respect to the vehicle 12100). In this way, the three-dimensional object that is the closest three-dimensional object on the path of the vehicle 12100 and is traveling at a predetermined speed (for example, 0 km / h or higher) in essentially the same direction as the vehicle 12100 can be identified as the vehicle traveling in front of the vehicle 12100.Furthermore, the microcomputer 12051 can predetermine a required distance between vehicles in front of the vehicle driving ahead of vehicle 12100, and can perform automatic braking control (including follow-stop control), automatic acceleration control (including follow-start control), and the like. In this way, it is possible to implement cooperative control for automated driving or similar functions, enabling autonomous driving without driver intervention. The microcomputer 12051 can, for example, extract three-dimensional object data based on distance information obtained from the imaging units 12101 to 12104. This data can be categorized into three-dimensional objects such as two-wheeled vehicles, ordinary vehicles, large vehicles, pedestrians, utility poles, and the like, and used for automatic obstacle avoidance. For instance, the microcomputer 12051 classifies obstacles near vehicle 12100 into those visible to the driver and those difficult to see. The microcomputer 12051 then determines collision risks, indicating the likelihood of a collision with each obstacle.If a collision risk is equal to or higher than a set value and there is a possibility of a collision, the microcomputer 12051 can issue a warning to a driver via the loudspeaker 12061 or the display unit 12062, or it can provide driving assistance to avoid a collision by initiating a forced braking or evasive steering movement via the drive system control unit 12010. At least one of the imaging units 12101 to 12104 can be an infrared camera that detects infrared radiation. For example, the microcomputer 12051 can detect a pedestrian by determining whether or not a pedestrian is present in the images captured by the imaging units 12101 to 12104. Such pedestrian detection is performed, for example, by a process to extract feature points from the images captured by the imaging units 12101 to 12104 acting as infrared cameras, and a process to perform a pattern match on the series of feature points that define the contours of objects and determine whether or not a pedestrian is present.If the microcomputer 12051 determines that a pedestrian is present in the images captured by the imaging units 12101 to 12104 and detects the pedestrian, the audio / video output unit 12052 controls the display unit 12062 so that the display unit 12062 overlays a rectangular outline to highlight the detected pedestrian. Furthermore, the audio / video output unit 12052 can also control the display unit 12062 to display a symbol or the like indicating a pedestrian at a desired position. In this description, "system" refers to a complete setup consisting of a large number of devices. Please note that the beneficial effects described in this description are only examples and the beneficial effects of the present technology are not limited to them or may include other effects. It should be noted that embodiments of the present technology are not limited to the embodiments described above and various modifications can be made to them without deviating from the scope of the present technology. Note that the present technology may also be embodied in the configurations described below. (1) A distance measuring device comprising: a first determination unit that determines whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; and a second determination unit that, when the first determination unit determines that the distance difference between the first pixel and the second pixel is greater than the first threshold, determines whether a confidence difference between the first pixel and the second pixel is greater than a second threshold, wherein, when the second determination unit determines that the confidence difference between the first pixel and the second pixel is greater than the second threshold, it confirms that the first pixel is a defective pixel. (2) The distance measuring device according to (1),(3) The distance measuring device according to (1) or (2), wherein the second threshold is set at a value obtained by multiplying an average confidence value of a plurality of second pixels by a predetermined value. (4) The distance measuring device according to (1) or (2), wherein the second threshold is set at a value obtained by multiplying a standard deviation of the confidence values ​​of a plurality of second pixels by a predetermined value. (5) The distance measuring device according to (1) or (2), wherein the second threshold is set at a value obtained by multiplying a difference between the largest and smallest confidence values ​​of a plurality of second pixels by a predetermined value. (6) The distance measuring device according to (1) or (2),wherein the second threshold is set at a value obtained by multiplying the confidence of the first pixel by a predetermined value.(7) A distance measuring device comprising: a first determination unit that determines whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; and a second determination unit that, when the first determination unit determines that the distance difference between the first pixel and the second pixel is greater than the first threshold, determines whether a reflectance difference between the first pixel and the second pixel is greater than a second threshold, wherein, when the second determination unit determines that the reflectance difference between the first pixel and the second pixel is greater than the second threshold, confirms,that the first pixel is a defective pixel. (8) The distance measuring device according to (7), wherein the reflectance is a value obtained by multiplying a confidence interval by the square of a depth value. (9) The distance measuring device according to (7) or (8), wherein the second threshold is a fixed value or a variable value. (10) The distance measuring device according to any one of (7) to (9), wherein the second threshold is set at a value obtained by multiplying an average reflectance of a plurality of the second pixels by a predetermined value. (11) The distance measuring device according to any one of (7) to (9), wherein the second threshold is set at a value obtained by multiplying a standard deviation of the reflectances of a plurality of the second pixels by a predetermined value. (12) The distance measuring device according to any one of (7) to (9),(13) The distance measuring device according to any one of (7) to (9), wherein the second threshold is set at a value obtained by multiplying the difference between the largest and smallest values ​​of the reflectances of a plurality of the second pixels by a predetermined value. (14) The distance measuring device according to any one of (7) to (9), wherein the second threshold is set at a value obtained by multiplying the reflectance of the first pixel by a predetermined value. (15) The distance measuring device according to any one of (7) to (9), wherein the second threshold is set to the sum of the square of a depth value of the first pixel and the square of a depth value of the second pixel. (16) A distance measuring method implemented by a distance measuring device that performs a distance measurement, wherein the distance measuring method comprises: determining,whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold or not; determining whether a confidence difference between the first pixel and the second pixel is greater than a second threshold or not, when it is determined that the distance difference between the first pixel and the second pixel is greater than the first threshold; and confirming that the first pixel is a defective pixel, when it is determined that the confidence difference between the first pixel and the second pixel is greater than the second threshold. (16) A distance measurement method implemented by a distance measuring device that performs a distance measurement, wherein the distance measurement method comprises: determining,whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether a reflectance difference between the first pixel and the second pixel is greater than a second threshold, if it is determined that the distance difference between the first pixel and the second pixel is greater than the first threshold; and confirming that the first pixel is a defective pixel, if it is determined that the reflectance difference between the first pixel and the second pixel is greater than the second threshold. (17) A program to cause a computer to perform a process comprising the steps: determining whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining,whether a confidence difference between the first pixel and the second pixel is greater than a second threshold or not, when it is determined that the distance difference between the first pixel and the second pixel is greater than the first threshold; and confirming that the first pixel is a defective pixel, when it is determined that the confidence difference between the first pixel and the second pixel is greater than the second threshold. (18) A program to cause a computer to perform a process comprising the steps of: determining whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold or not; determining whether a reflectance difference between the first pixel and the second pixel is greater than a second threshold or not, when it is determined thatthat the distance difference between the first pixel and the second pixel is greater than the first threshold; and confirming that the first pixel is a defective pixel when it is determined that the reflectance difference between the first pixel and the second pixel is greater than the second threshold. (19) An electronic device comprising a processing unit that performs a process using a machine-learned learning model on at least part of a first distance measurement information acquired by a sensor and outputs a second distance measurement information after a correction target pixel contained in the first distance measurement information has been corrected, wherein the process comprises: a first process to correct the correction target pixel using the first distance measurement information containing the correction target pixel as input; and a second process,to output the second distance measurement information. (20) The electronic device according to (19), wherein the first process includes a first step to identify the correction target pixel. (21) The electronic device according to (20), wherein the first process includes a second step to correct the identified correction target pixel. (22) The electronic device according to (21), wherein a process utilizing the learning model is performed in the first step or the second step. (23) The electronic device according to (21), wherein a process utilizing the learning model is performed in the first step and the second step. (24) The electronic device according to any one of (20) to (23), wherein the first distance measurement information is an uncorrected depth map and the first step comprises: Determining,whether a distance difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; Determine whether a confidence or reflectance difference between the first pixel and the second pixel is greater than a second threshold, if it is determined that the distance difference between the first pixel and the second pixel is greater than the first threshold; and Confirm that the first pixel is the correction target pixel, if it is determined that the confidence or reflectance difference between the first pixel and the second pixel is greater than the second threshold. (25) The electronic device according to (19), wherein the process further includes a third process to generate three-dimensional data based on the second distance measurement information. (26) The electronic device according to (25),(27) The electronic device according to (26), wherein the first to the third processes are carried out in a continuous machine learning process. (28) The electronic device according to any one of (19) to (27), wherein the first distance measurement information is an uncorrected depth map and the second distance measurement information is a corrected depth map. (29) The electronic device according to any one of (19) to (28), wherein the correction target pixel is a faulty pixel or a floating pixel. (30) The electronic device according to any one of (19) to (29), wherein the learning model comprises a neural network that has learned from a dataset to identify the correction target pixel. (31) The electronic device according to any one of (20) to (23),wherein the initial distance measurement information is an uncorrected depth map and the first step comprises the steps of: inputting a depth map containing the correction target pixel into an input layer of a neural network using the neural network that has learned a normal depth map with an autocoder, and outputting a depth map in which the correction target pixel has been corrected from an output layer of the neural network; and identifying the correction target pixel based on the depth map input into the input layer and position information in the depth map output by the output layer. (32) The electronic device according to (31),(33) The electronic device according to (28), wherein the first process includes the step of inputting a depth map containing the correction target pixel into an input layer of a neural network and outputting a depth map in which the correction target pixel has been corrected from an output layer of the neural network, using the neural network that has learned a normal depth map with an autocoder. (34) The electronic device according to any one of (19) to (33), further comprising the sensor, wherein the sensor contains the processing unit. (35) The electronic device according to any one of (19) to (33),wherein the electronic device is configured as a mobile terminal or server.(36) A method for generating a training model, comprising: acquiring training data containing a depth map that is a depth map containing a correction target pixel, wherein a position of the correction target pixel in the depth map has been identified; and generating a training model using the training data, wherein the training model receives an input of the depth map containing the correction target pixel and outputs position information about the correction target pixel contained in the depth map.(37) The method for generating a training model according to (36), wherein the position information includes information relating to the coordinates of the correction target pixel or information relating to an anomaly probability of each pixel.(38) The method for generating a training model according to (36) or (37), wherein, when the training model is generated,The positional information about the correction target pixel is acquired by: determining whether a distance difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether a confidence or reflectance difference between the first pixel and the second pixel is greater than a second threshold, if it is determined that the distance difference between the first pixel and the second pixel is greater than the first threshold; and confirming that the first pixel is the correction target pixel, if it is determined that the confidence or reflectance difference between the first pixel and the second pixel is greater than the second threshold. (39) A manufacturing process that includes the step of manufacturing an electronic device,(40) A manufacturing process comprising the step of manufacturing an electronic device storing the learning model by storing the learning model generated by the learning model generating process according to (36) to (38) on a readable recording medium. (41) A depth map generating process comprising the steps of: inputting distance measurement information acquired by a sensor into the learning model generated by the learning model generating process according to (36) to (38); and generating a depth map as output of the learning model, wherein a correction target pixel in the depth map has been corrected. (42) A manufacturing process comprising the step of manufacturing an electronic device storing a corrected depth map by storing the corrected depth map,(43) A method for generating a training model, comprising: acquiring training data containing a depth map including a correction target pixel and position information about the correction target pixel and a corrected depth map or position information about a corrected correction target pixel; and generating a training model with the training data, wherein the training model receives input of the depth map containing the correction target pixel and the position information about the correction target pixel and outputs the corrected depth map or the position information about the corrected correction target pixel. (44) The method for generating a training model according to (43), wherein the position information includes information relating to the coordinates of the correction target pixel. (45) A manufacturing method,(46) A manufacturing process that includes the step of manufacturing an electronic device in which the learning model generated by the learning model generation process according to (43) or (44) is installed. (47) A manufacturing process that includes the step of manufacturing an electronic device storing the learning model by storing the learning model generated by the learning model generation process according to (43) or (44) on a readable recording medium. (48) A depth map generation process that includes the steps of: inputting distance measurement information acquired by a sensor into the learning model generated by the learning model generation process according to (43) to (44); and generating a depth map as output of the learning model, wherein a correction target pixel in the depth map has been corrected. (49) A manufacturing process that includes the step of manufacturing an electronic device storing a corrected depth map.by storing the corrected depth map produced by the depth map-generating method according to (47) on a readable recording medium. (49) A depth map-generating method implemented by a distance-measuring device performing a distance measurement, wherein the depth map-generating method comprises: determining whether a distance difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether a confidence or reflectance difference between the first pixel and the second pixel is greater than a second threshold when it is determined that the distance difference between the first pixel and the second pixel is greater than the first threshold; confirming that the first pixel is a correction target pixel when it is determined thatthat the confidence or reflectance difference between the first pixel and the second pixel is greater than the second threshold; and the generation of a depth map in which the confirmed correction target pixel has been corrected. REFERENCE MARK LIST 10 Distance measuring device 11 Lens 12 Light receiving unit 13 Signal processing unit 14 Light emitting unit 15 Light emission control unit 16 Filter unit 31 Photodiode 41 Pixel array unit 42 Vertical drive unit 43 Column processing unit 44 Horizontal drive unit 45 System control unit 46 Pixel drive line 47 Vertical signal line 48 Signal processing unit 50 Pixel 51 Tap 61 Photodiode 62 Transmission transistor 63 FD area 64 Reset transistor 65 Gain transistor 66 Selection transistor 101 Foreground object 102 Background object 103, 104, 105 Border 111, 112 Dotted line 121 Pixel group 122 Pixel group 123, 124 Pixel 150, 152, 154, 156, 158 pixels 601 electronic device 602 sensor 603 controller 604 output unit 611 distance measuring unit 612 processing unit 621 learning model 701 learning device 703 generation unit 704 output unit< / aufzeichnungsmedium>

Claims

Distance measuring device comprising: a first determination unit that determines whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; and a second determination unit that, if the first determination unit determines that a distance difference between the first pixel and the second pixel is greater than the first threshold, determines whether a confidence difference between the first pixel and the second pixel is greater than a second threshold, wherein, if the second determination unit determines that the confidence difference between the first pixel and the second pixel is greater than the second threshold, it confirms that the first pixel is a defective pixel. Distance measuring device claim 1, wherein the second threshold is a fixed value or a variable value. Distance measuring device according to claim 1, wherein the second threshold is set at a value obtained by multiplying an average confidence value of a plurality of second pixels by a predetermined value. Distance measuring device according to claim 1, wherein the second threshold is set at a value obtained by multiplying a standard deviation of the confidences of a plurality of the second pixels by a predetermined value. Distance measuring device according to claim 1, wherein the second threshold is set at a value obtained by multiplying a difference between a largest value and a smallest value of the confidences of a plurality of the second pixels by a predetermined value. Distance measuring device according to claim 1, wherein the second threshold is set at a value obtained by multiplying the confidence of the first pixel by a predetermined value. Distance measuring device comprising: a first determination unit that determines whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; and a second determination unit that, if the first determination unit determines that a distance difference between the first pixel and the second pixel is greater than the first threshold, determines whether a reflectance difference between the first pixel and the second pixel is greater than a second threshold, wherein, if the second determination unit determines that the reflectance difference between the first pixel and the second pixel is greater than the second threshold, it confirms that the first pixel is a defective pixel. Distance measuring device according to claim 7, wherein the reflectance is a value obtained by multiplying a confidence interval by the square of a depth value. Distance measuring device according to claim 7, wherein the second threshold is a fixed value or a variable value. Distance measuring device according to claim 7, wherein the second threshold is set at a value obtained by multiplying an average value of the reflectances of a plurality of the second pixels by a predetermined value. Distance measuring device according to claim 7, wherein the second threshold is set at a value obtained by multiplying a standard deviation of the reflectances of a plurality of the second pixels by a predetermined value. Distance measuring device according to claim 7, wherein the second threshold is set at a value obtained by multiplying a difference between a largest value and a smallest value of the reflectances of a plurality of the second pixels by a predetermined value. Distance measuring device according to claim 7, wherein the second threshold is set at a value obtained by multiplying the reflectance of the first pixel by a predetermined value. Distance measuring device according to claim 7, wherein the second threshold is set to a sum of a square of a depth value of the first pixel and a square of a depth value of the second pixel. A distance measurement method implemented by a distance measurement device that performs a distance measurement, wherein the distance measurement method comprises: determining whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether a confidence difference between the first pixel and the second pixel is greater than a second threshold when it is determined that a distance difference between the first pixel and the second pixel is greater than the first threshold; and confirming that the first pixel is a defective pixel when it is determined that the confidence difference between the first pixel and the second pixel is greater than the second threshold. A distance measurement method implemented by a distance measurement device that performs a distance measurement, wherein the distance measurement method comprises: determining whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether a reflectance difference between the first pixel and the second pixel is greater than a second threshold when it is determined that a distance difference between the first pixel and the second pixel is greater than the first threshold; and confirming that the first pixel is a defective pixel when it is determined that the reflectance difference between the first pixel and the second pixel is greater than the second threshold. A program to cause a computer to perform a process that includes the following steps: Determine whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; Determine whether a confidence difference between the first pixel and the second pixel is greater than a second threshold if a distance difference between the first pixel and the second pixel is determined to be greater than the first threshold; and Confirm that the first pixel is a faulty pixel if the confidence difference between the first pixel and the second pixel is determined to be greater than the second threshold. A program to cause a computer to perform a process that includes the following steps: Determine whether a depth value difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; Determine whether a reflectance difference between the first pixel and the second pixel is greater than a second threshold if a distance difference between the first pixel and the second pixel is determined to be greater than the first threshold; and Confirm that the first pixel is a faulty pixel if the reflectance difference between the first pixel and the second pixel is determined to be greater than the second threshold. A depth map generating method implemented by a distance measuring device performing a distance measurement, wherein the depth map generating method comprises: determining whether a distance difference between a first pixel in a depth map and a second pixel adjacent to the first pixel is greater than a first threshold; determining whether a confidence or reflectance difference between the first pixel and the second pixel is greater than a second threshold when it is determined that the distance difference between the first pixel and the second pixel is greater than the first threshold; confirming that the first pixel is a correction target pixel when it is determined that the confidence or reflectance difference between the first pixel and the second pixel is greater than the second threshold; and generating a depth map wherein the confirmed correction target pixel has been corrected.

Images