Distance measurement control device, distance measurement control program, distance measurement method, robot

Abstract

Accurate distance information is obtained from images captured with a monocular camera. [Solution] For each captured image, the relative depth values ​​of the positions indicated by two reference points A and B, whose distances are known, are read out. A characteristic line is generated showing the relationship between the inferred value (relative depth value) and the measured value (distance), which is directly proportional. Based on this characteristic line, the distance information of the point cloud data is corrected. As a result, within the captured image, the point cloud data accurately represents distances based on the characteristic line.

Description

[Technical Field]

[0001] The present invention relates to a distance measurement control device, a distance measurement control program, a distance measurement method, and a robot for measuring distance information of each point in an image captured by a monocular camera. [Background Art]

[0002] For example, when an actuator such as an arm is attached to a robot to automatically perform work, the operation of the actuator (moving direction, moving distance, etc.) and the position of a target object (three-dimensional coordinates including distance) are required.

[0003] Patent Document 1 describes that teaching of a movement path of a robot arm (actuator) in real space is easily performed. More specifically, a robot control system includes a robot arm, a three-dimensional camera attached to the robot arm, and a robot control device. The robot control device includes a control unit. The control unit acquires two-dimensional image data and point cloud data of a target object by the three-dimensional camera, recognizes a reference line on the surface of the target object based on the two-dimensional image data, and controls movement of the robot arm to move the three-dimensional camera along the reference line. [Prior Art Literature] [Patent Literature]

[0004] [Patent Document 1] Japanese Unexamined Patent Application Publication No. 2023-004158 [Summary of the Invention] [Problem to be Solved by the Invention]

[0005] In order to obtain the operation of an actuator, the distance to a target object, and the like, point cloud data acquirable by capturing an image with a 3D camera is required. However, the introduction of a 3D camera may be restricted depending on cost, installation environment, and other factors.

[0006] Therefore, robots equipped with a monocular camera (2D camera) to monitor their surroundings are sometimes employed. In this case, the images captured by the 2D camera are analyzed using deep learning processing in the field of monocular vibration estimation to estimate depth information.

[0007] However, this depth information obtains an 8-bit (0-255) relative depth value for images with different depth ranges. The depth range is the difference between the minimum and maximum depths (dynamic range), and because the dynamic range varies for each captured image, and the distance (resolution) per unit scale (1 bit) differs, it is not always possible to obtain an accurate distance.

[0008] Considering the above facts, the present invention aims to provide a distance measurement control device, a distance measurement control program, a distance measurement method, and a robot that can obtain distance information with minimal error from images captured by a monocular camera. [Means for solving the problem]

[0009] The distance measurement control device according to the present invention includes an estimation unit that captures an image by a monocular camera to include a plurality of reference points set at different predetermined distances, and converts the image into a point cloud of predetermined density, and estimates the relative depth value of each point cloud with respect to an object to be measured that is present in the shooting range; and a calculation unit that calculates the distance based on the correlation between the relative depth value of each of the plurality of reference points whose distances are known and which are set in advance, and the respective distances, from among the point cloud estimated by the estimation unit.

[0010] The distance measurement control device according to the present invention includes an estimation unit that captures an image by a monocular camera to include a plurality of reference points set at different predetermined distances, and converts the image into a point cloud of a predetermined density, and estimates the relative depth value of each point cloud with respect to an object to be measured that is present in the shooting range; and a correction unit that corrects the relative depth value based on the relative depth value of each of a plurality of reference points among the point cloud estimated by the estimation unit, each of which has a known reference relative depth value.

[0011] The distance measurement control program according to the present invention is characterized by operating a computer as the distance measurement control device described above.

[0012] The distance measurement method according to the present invention is a distance measurement method for a distance measurement control device that measures distance information for each point by converting an image captured by a monocular camera into a point cloud of predetermined density, wherein the distance measurement control device includes a first step of converting an image captured by a monocular camera into a point cloud of predetermined density so as to include a plurality of reference points set at different predetermined distances, and estimating the relative depth value of each point cloud with respect to an object to be measured that exists in the shooting range, and a second step of calculating the distance based on the correlation between the relative depth value of each of a plurality of reference points whose distances are known and which are set in advance, and the respective distances among the estimated point cloud.

[0013] The distance measurement method according to the present invention is a distance measurement method for a distance measurement control device that measures distance information for each point by converting an image captured by a monocular camera into a point cloud of predetermined density, wherein the distance measurement control device includes a first step of converting an image captured by a monocular camera into a point cloud of predetermined density so as to include a plurality of reference points set at different predetermined distances, and estimating the relative depth value of each point cloud with respect to an object to be measured that exists in the shooting range, and a second step of correcting the relative depth value based on the relative depth value of each of a plurality of reference points among the estimated point cloud whose reference relative depth values ​​are known.

[0014] The robot according to the present invention includes an estimation unit that captures images taken by a monocular camera to include a plurality of reference points set at different predetermined distances, converts them into a point cloud of predetermined density, and estimates the relative depth value of each point cloud with respect to an object to be measured that exists within the shooting range; a calculation unit that calculates the distance based on the correlation between the relative depth value of each of a plurality of reference points whose distances are known and which are set in advance, and the respective distances; and a conversion unit that converts the relative depth value of a predetermined part of the object to be measured into three-dimensional position coordinates based on the distance calculated by the calculation unit; and a control unit that operates an actuator according to the distance obtained by the distance measurement control unit to control the work.

[0015] The robot according to the present invention includes an estimation unit that captures images taken by a monocular camera to include a plurality of reference points set at different predetermined distances, and estimates the relative depth of each point in the point cloud with respect to an object to be measured that is present in the shooting range; a correction unit that corrects the relative depth of each of the plurality of reference points within the point cloud estimated by the estimation unit based on the relative depth of each of the plurality of reference points whose reference relative depth is known and each of the reference relative depths; a conversion unit that converts the relative depth of a predetermined part of the object to be measured into three-dimensional position coordinates based on the relative depth of the corrected relative depth of the correction unit; and a control unit that operates an actuator according to the distance obtained by the distance measurement control unit to control the work. [Effects of the Invention]

[0016] As described above, the present invention makes it possible to obtain distance information with minimal error from images captured by a monocular camera. [Brief explanation of the drawing]

[0017] [Figure 1] (A) is a perspective view of a robot equipped with a distance measuring control device according to this embodiment, and (B) is a control block diagram showing the control system for operating the robot. [Figure 2] (A) is a side view of the robot according to this embodiment, (B) is a top view of the robot according to this embodiment, and (C) is a functional block diagram for distance measurement control performed by the distance measurement control device. [Figure 3] This is a characteristic curve of measured value (distance) - inferred value (relative depth value) according to this embodiment. [Figure 4] This is a control flowchart showing the flow of distance measurement control performed by the distance measurement control device according to this embodiment. [Figure 5] This is an embodiment of this design, where (A) is a front view of an image captured by the camera unit, and (B) is a front view of the point cloud data of the image in Figure 5(A). [Modes for carrying out the invention]

[0018] Figs. 1(A) and 1(B) are perspective views of a robot 12 equipped with a distance measurement control device 10 (see Fig. 1(B)) according to the present embodiment. Note that the structure of the robot 12 shown in Figs. 1(A) and 1(B) is an example, and does not limit the present invention.

[0019] The robot 12 is assembled with a first housing 12A representing a head, a second housing 12B representing a torso, and a third housing 12C representing feet, and although not shown in the figures, is provided with actuators such as arms necessary for work.

[0020] A camera unit 16 is attached to the first housing 12A. The camera unit 16 has a monocular camera structure, and mainly captures images of the area ahead of the robot 12 in the movement direction. A rotatable shaft portion 12D is provided between the first housing 12A and the second housing 12B, so that the optical axis of the camera unit 16 can be redirected vertically and horizontally.

[0021] The second housing 12B includes the distance measurement control device 10 and a robot control unit 18 (see Fig. 1(B)).

[0022] The distance measurement control device 10 has a function of generating point cloud data from an image captured by the camera unit 16 and acquiring distance information to, for example, a measurement object 19 shown in Fig. 1(A), and details thereof will be described later.

[0023] As shown in Fig. 1(B), the distance measurement control device 10 includes a microcomputer 20. The microcomputer 20 has a CPU (Central Processing Unit) 20A, a RAM (Random Access Memory) 20B, a ROM (Read Only Memory) 20C, an input / output device (I / O) 20D, and a bus 20E such as a data bus or a control bus connecting these components.

[0024] A hard disk 23 is connected to I / O20D. Additionally, a camera unit 16 of the first enclosure 12A is connected to I / O20D via an interface 24. Furthermore, a robot control unit 18 is connected to I / O20D.

[0025] The robot control unit 18 controls the operation of actuators to perform tasks based on a predetermined program. Specifically, the robot 12 moves to a predetermined position, controls the operation of multiple drive sources (drive source group 14B) based on detection signals from multiple sensors (sensor group A), and performs predetermined tasks using actuators (not shown).

[0026] Furthermore, instead of using a pre-programmed system, artificial intelligence such as AI may be used to independently determine the work format and operate based on the operator's prompts (information that facilitates the input of command prompts, which are instructions to the computer).

[0027] The third housing 12C is equipped with a mobile unit 26 for the robot 12 to move. Although not shown in the figure, the mobile unit 26 is equipped with a motor as a drive source and wheels that rotate due to the motor's drive, and operates based on instructions from the robot control unit 18 to move the robot 12. Note that movement is not limited to wheels; it may also be controlled by walking.

[0028] As shown in Figure 1(A) and Figures 2(A) and (B), in this embodiment, the robot 12 has a pair of rods 28 and 30 extending from the second housing 12B in the direction of forward movement of the robot 12.

[0029] The rods 28 and 30 have different lengths, with the relationship L1 of one rod 28 being less than that of the other rod 30.

[0030] Poles 28A and 30A are erected at the tips of each of the rod bodies 28 and 30, with their upper ends designated as reference points A and B, respectively, and are always located within the shooting range when the camera unit 16 is taking pictures.

[0031] A pair of rods 28 and 30 are attached to the base of the second housing 12B by triangular support members 28B and 30B, which limit the vibration of the pair of rods 28 and 30 caused by the movement of the robot 12. This makes it possible to suppress fluctuations of reference points A and B. Note that vibration suppression is not limited to support members 28B and 30B, but may also be achieved by a double-structured rod.

[0032] Furthermore, the structural parts where reference points A and B are located are placed outside the operating range of the actuator so as not to interfere with its operation.

[0033] Although reference points A and B are positioned so that their distance from each other is fixed, they may also be attached to actuators or the like. In this case, reference points A and B will fluctuate due to the operation of the actuator, but the distance between reference points A and B can be determined based on the actuator's control signal.

[0034] Here, the bases of the rods 28 and 30 are attachment points to the second housing 12B, and these attachment points are the origin position when the camera unit 16 takes a picture. The distances from this origin position (shooting distance = 0 point) to reference point A and reference point B are known. In this embodiment, for example, the length L1 from the origin position to reference point A is 123 cm, and the length L2 from the origin position to reference point B is 175 cm.

[0035] In other words, the distances to reference point A and reference point B in the images captured by camera unit 16 are known (L1 = 123 cm, L2 = 175 cm).

[0036] The distance measurement control device 10 generates point cloud data from images captured by the camera unit 16, and the point cloud data can represent the distance (depth information) of each part of the shooting range using 8 bits (0 to 255).

[0037] Here, the relative depth value represents the relative relationship of depth within the captured image range. In the comparative example, the distance to each part within the shooting range was obtained as an inferred value based on analysis of the shooting location, etc., from this relative depth value.

[0038] In contrast, in this embodiment, since the aforementioned reference points A and B are photographed and the distance between reference points A and B is known, the relationship between the relative depth value and distance at the positions corresponding to reference points A and B is obtained.

[0039] (Details of the relationship between relative depth and distance) In this embodiment, since the camera unit 16 is a monocular camera, it is not possible to obtain the precise distance to each position within the shooting range, as is the case with 3D cameras. Therefore, it is necessary to estimate depth information based on deep learning, which is well known in fields such as monocular depth estimation.

[0040] Here, the estimated depth information differs depending on the image being captured, with different minimum and maximum depths (different depth ranges).

[0041] For images with different depth ranges, an 8-bit (0-255) relative depth value is assigned.

[0042] In other words, the dynamic range varies depending on the image captured, and the distance (resolution) per unit scale (1 bit unit) differs.

[0043] Based on this relative depth value, the captured image is converted into point cloud data containing distance information (x, y, z coordinates).

[0044] However, even if the relative depth value is the same (for example, 128), the distance to that position will differ for each captured image.

[0045] Therefore, for each captured image, the relative depth values ​​of the positions indicated by two reference points A and B, whose distances are known, are read out, and a characteristic line is generated that shows the relationship between the inferred value (relative depth value) and the measured value (distance), which is in a directly proportional relationship.

[0046] Since the inferred value and the measured value are directly proportional, the generated characteristic line is a straight line graph with an intercept of 255 and a negative slope a, as shown in Figure 3. Note that the general formula is written as y = ax + b, but the variables x and y here are different from the coordinates x, y, and z of the distance information mentioned earlier.

[0047] Based on this characteristic curve, the distance information of the point cloud data is corrected.

[0048] As a result, within the captured image, the point cloud data represents accurate distances based on characteristic lines.

[0049] It is assumed that the captured images are video, and it is preferable to create characteristic lines for each frame, but depending on the processing capabilities such as processing speed, characteristic lines may be created for every multiple frames.

[0050] Furthermore, since video rarely experiences extreme changes in depth (fluctuations in dynamic range), it may be acceptable to create characteristic curves only when the dynamic range clearly changes.

[0051] Fluctuations in the dynamic range can be detected when the position of the image with the minimum and maximum depth values ​​moves by more than a predetermined amount. Alternatively, to put it extremely, assuming that the working environment remains virtually unchanged, the system could generate a characteristic line for correction only once when it starts up.

[0052] Figure 2(B) is a control block diagram of the distance measurement control device 10, classifying the control functions for generating a measured value (distance)-inferred value (relative depth value) characteristic line from captured images and correcting point cloud data. The functions of each block are operated by the microcomputer shown in Figure 1(A) based on a software program. IC chips (semiconductor integrated circuits) such as ASICs (Application Specific Integrated Circuits), FPGAs (Field Programmable Gate Arrays), and CPLDs (Complex Programmable Logic Devices) may be used to operate some or all of the functions of each block.

[0053] The camera unit 16 is connected to the image acquisition unit 50 and acquires captured images. The image acquisition unit 50 is connected to the depth information estimation unit 52 and transmits the acquired images.

[0054] The depth information estimation unit 52 performs depth estimation within the range of the received image. The estimated depth is represented by 8 bits, with the minimum and maximum depths within the range of the image used as the dynamic range. In other words, the depth information estimation unit 52 performs depth information estimation processing based on deep learning on images captured by a monocular camera, and does not directly acquire distance information.

[0055] The depth information estimation unit 52 sends the data to the point cloud data conversion unit 54, which converts the captured image into point cloud data based on the estimated depth information. Point cloud data is a set of coordinates (x, y, z) of observation points on the surface of an object in a three-dimensional space, where the captured image is treated as a three-dimensional space. At this point, it is generated based on the estimated depth information.

[0056] When the estimated depth information is converted into point cloud data by the point cloud data conversion unit 54, the point cloud data is sent to the reference value position depth relative value acquisition unit 56.

[0057] The reference value position depth relative value acquisition unit 56 identifies the positions of reference points A (distance L1) and B (distance L2), whose distances are known and attached to the robot 12, reads out the depth (depth relative value) estimated at the identified positions, and sends it to the characteristic line generation unit 58.

[0058] The characteristic line generation unit 58 generates a measured value (distance) - inferred value (relative depth value) characteristic line (see Figure 3) for the image, based on the fact that the relationship between the relative depth value and distance is directly proportional (y=ax+b), and the relationship between the relative depth values ​​and distances of two points (reference point A and reference point B).

[0059] The characteristic line generation unit 58 is connected to the point cloud data correction unit 60. The characteristic line generation unit 58 sends the generated measured value (distance) - inferred value (depth relative value) characteristic line to the point cloud data correction unit 60.

[0060] The point cloud data correction unit 60 performs corrections on the distance-related information among the coordinates (x, y, z) transformed by the point cloud data conversion unit 54, based on the measured value (distance) - inferred value (depth relative value) characteristic line.

[0061] More specifically, robot 12 is configured to determine distance from reference point cloud data (reference image). However, errors occur when determining distance using depth estimates from actual measurement images.

[0062] Therefore, depth characteristic curves are used to correct the depth estimate of the measured image to the depth estimate of the reference image, which is the basis for moving the robot.

[0063] As a result, robot 12 can obtain depth estimates based on the criteria for moving the robot.

[0064] The point cloud data corrected by the point cloud data correction unit 60 has distance data that is equivalent to, or comparable to, point cloud data generated by, for example, a 3D camera, and is sent to the robot control unit 18 via the output unit 62.

[0065] The operation of this embodiment will be explained below in accordance with the flowchart in Figure 4.

[0066] The flowchart shown in Figure 4 is executed by the distance measurement control device 10, and processing begins, for example, when the robot 12 (robot control unit 18) system is started up.

[0067] In step 100, shooting begins with the camera unit 16, and then the process proceeds to step 102 to determine whether or not an image has been acquired.

[0068] If the result in step 102 is negative, the process proceeds to step 116 (the processing from step 116 onwards is described later). If the result in step 102 is positive, the process proceeds to step 104 to execute depth estimation. Depth estimation estimates the dynamic range between the minimum and maximum depth of the image and represents it as an 8-bit (0-255) relative depth value. Note that the relative depth value is not limited to 8 bits; it may also be 16 bits, such as -255 to +255. Furthermore, it is not necessary to stick to binary (bit units); decimal or hexadecimal values ​​may also be used. For example, in decimal, it may be represented as a number between -1 and +1 (decimal point), or as a percentage.

[0069] In the next step, 106, the relative depth values ​​are converted into point cloud data containing distance information, and the process moves on to step 108.

[0070] In step 108, the relative depth values ​​of the point cloud data located at reference points A and B are obtained, and then the process moves to step 110 to generate a measured value (distance) - inferred value (relative depth value) characteristic line (y=ax+b) (see Figure 3).

[0071] In the next step 112, the distance information of the point cloud data is corrected based on the characteristic lines generated in step 110. Then, the process moves to step 114, where the corrected point cloud data is transmitted to the robot control unit 18, and the process moves to step 116.

[0072] In step 116, it is determined whether the system has terminated or not. If the determination is negative, the process returns to step 100 and the above steps are repeated. If the determination in step 116 is positive, the process proceeds to step 118 to terminate the recording, and this routine ends.

[0073] As described above, in this embodiment, for each captured image, the relative depth values ​​of the positions indicated by two reference points A and B, whose distances are known, are read out, and a characteristic line is generated that shows the relationship between the inferred value (relative depth value) and the measured value (distance), which is in a directly proportional relationship. Based on this characteristic line, the distance information of the point cloud data is corrected. As a result, within the captured image, the point cloud data represents accurate distances based on the characteristic line.

[0074] In this embodiment, first, the data is converted into point cloud data including three-dimensional coordinates based on the relative depth values ​​(0 to 255) for each point of a predetermined density obtained from the captured image. Then, the distance information of the point cloud data is corrected based on the measured value (distance) - inferred value (relative depth value) characteristic line in Figure 3. However, the relative depth values ​​may be corrected based on the measured value (distance) - inferred value (relative depth value) characteristic line in Figure 3 before converting to point cloud data, and then converted to point cloud data. Alternatively, instead of the measured value (distance) - inferred value (relative depth value) characteristic line, a reference value - inferred value (relative depth value) may be used, and the data may be directly converted to a reference value used for robot operation without calculating the distance.

[0075] (Examples) Figures 5(A) and (B) show an example of the process of correcting point cloud data (estimated depth) with characteristic lines when the distance measurement control device 10 according to this embodiment is applied.

[0076] Figure 5(A) is an image taken with the object to be measured 19 positioned within the shooting range of the camera unit 16 in front of the robot 12, as shown in Figure 1(A). While a human (measurer) can visually see that the object is at a greater distance than reference point B, the 8-bit depth values ​​obtained from depth estimation were estimated to be 242 for reference point A and 175 for reference point B. Similarly, depth estimates are calculated for the object to be measured 19 and other elements within the image, and point cloud data is generated.

[0077] Here, based on the characteristic line (see Figure 3) generated using the known distances L1 (123 cm) of reference point A and L2 (175 cm) of reference point B, the distance corresponding to the depth estimate (8 bits) of the object being measured 19 is read. The read distance is an accurate value equivalent to or comparable to that obtained by capturing images with a 3D camera and generating point cloud data.

[0078] The calculation procedure using Figure 3 (x: inferred value (relative depth value), y: measured value (distance)) is shown below. As shown in Figure 3, the xy characteristic curves are directly proportional, so the equation of the characteristic curve is expressed as follows. y = ax + b ... (1) Reference point A and reference point B have the relationship shown in Table 1.

[0079] [Table 1]

[0080] Substitute the values ​​from Table 1 into x and y in (1) and perform the calculation to find the coefficients a and b. 175.0 = 163.0a + b ... (2) 242.0 = 123.0a + b ... (3) From equation (2), b = 175.0 - 163.0a ···(4) Substituting equation (4) for b in equation (3), 242.0=123.0a+(175.0-163.0a)...(5) (5) We can find the coefficient a from equation (5). -123.0a + 163.0a = 175.0 - 242.0 40a = -67 a = -(67 / 40) = -1.675 ... (6) Substitute equation (6) into equation (4) to find the coefficient b. b=175.0-(163.0×-1.675)=175.0-(-273.025)≒448 Therefore, equation (1) is, y = -1.675x + 448 ... (7) This is the result. Here, for example, if the inferred value y of the object to be measured 19, as shown in Figure 5(B), is 168.0, the measured value x is calculated using equation (7). 168.0 = -1.675x + 448 168 - 448 = -1.675x -280 = -1.675x x = 280 / 1.675 ≈ 167.2 Therefore, the measured value x of object 19 is 167.2 cm. In this way, the point cloud data of the position of the object to be measured 19 is corrected using equation (7) above (characteristic diagram in Figure 3).

[0081] Similar to the object 19 being measured, by correcting the positional information of each point cloud data within the image, it is possible to obtain point cloud data with accurate distance information for the entire image. [Explanation of Symbols]

[0082] A, B reference point 10 Distance measuring control device 12 Robots 12A First enclosure 12B Second cabinet 12C Third Cabinet 12D shaft part 14A Sensor Group 14B Drive Source Group 16 Camera Unit 18 Robot Control Unit 19. Object to be measured 20 Microcomputers 20A CPU 20B RAM 20C ROM 20D Input / output device (I / O) 20E Bus 23 Hard Disk 24 I / F 26 Mobile Units 28, 30 Rods 28A, 30A pole 50 Image acquisition unit 52 Depth information estimation section (estimation section) 54 Point Cloud Data Conversion Unit (Conversion Unit) 56 Reference value position depth relative value acquisition unit 58 Characteristic Line Generation Unit (Calculation Unit) 60-point cloud data correction unit (correction unit) 62 Output section

Claims

1. An estimation unit that captures images taken by a monocular camera, including multiple reference points set at different predetermined distances, into a point cloud of predetermined density, and estimates the relative depth value of each point cloud with respect to the object to be measured within the shooting range, A calculation unit calculates the distance based on the correlation between the relative depth value of each of several reference points with known distances, which are set in advance, and the respective distances, among the point cloud estimated by the estimation unit. A distance measuring device having

2. An estimation unit that captures images taken by a monocular camera, including multiple reference points set at different predetermined distances, into a point cloud of predetermined density, and estimates the relative depth value of each point cloud with respect to the object to be measured within the shooting range, The estimation unit estimates the point cloud, and each of the reference points has known reference depth relative values, and the correction unit corrects the depth relative values ​​based on each of the reference depth relative values. A distance measuring device having

3. The distance measuring control device according to claim 1, further comprising a conversion unit that converts the depth relative value of a predetermined part of the object to be measured into three-dimensional position coordinates based on the distance calculated by the calculation unit.

4. The distance measuring control device according to claim 2, further comprising a conversion unit that converts the relative depth values ​​of a predetermined part of the object to be measured into three-dimensional position coordinates based on the relative depth values ​​corrected by the correction unit.

5. The distance measuring control device according to claim 1 or claim 2, wherein the correlation between the relative depth value and the distance is in a direct proportional relationship, and two points on the characteristic line of the direct proportional relationship are known by the relative depth values ​​of each of the two reference points at different distances, thereby obtaining a correlation characteristic line.

6. Computers, To be operated as a distance measuring control device according to claim 1 or claim 2, Distance measurement control program.

7. A distance measurement method for a distance measurement control device that converts images captured by a monocular camera into a point cloud of predetermined density and measures the distance information of each point, The distance measuring control device is The first step involves capturing images with a monocular camera to form a point cloud of a predetermined density, including multiple reference points set at different predetermined distances, and estimating the relative depth value of each point cloud with respect to the object to be measured within the shooting range. A second step involves calculating the distance based on the correlation between the relative depth values ​​of each of several reference points with known distances, which are pre-set, and their respective distances, from among the estimated point cloud. Distance measurement methods including

8. A distance measurement method for a distance measurement control device that converts images captured by a monocular camera into a point cloud of predetermined density and measures the distance information of each point, The distance measuring control device is The first step involves capturing images with a monocular camera to form a point cloud of a predetermined density, including multiple reference points set at different predetermined distances, and estimating the relative depth value of each point cloud with respect to the object to be measured within the shooting range. A second step involves correcting the relative depth values ​​based on the relative depth values ​​of each of several reference points with known reference depth relative values ​​among the estimated point cloud, Distance measurement methods including

9. The distance measurement method according to claim 7, further comprising a third step of converting the depth relative values ​​of a predetermined part of the object to be measured into three-dimensional position coordinates based on the calculated distance after the execution of the second step.

10. The distance measurement method according to claim 8, further comprising a third step of converting the depth relative values ​​of a predetermined part of the object to be measured into three-dimensional position coordinates based on the corrected depth relative values ​​after the execution of the second step.

11. A distance measurement control device comprising: an estimation unit that captures images by a monocular camera so as to include multiple reference points set at different predetermined distances, converts them into a point cloud of predetermined density, and estimates the relative depth value of each point cloud for an object to be measured that exists within the shooting range; a calculation unit that calculates the distance based on the correlation between the relative depth value of each of the multiple reference points whose distances are predetermined and known, and the respective distances, and a conversion unit that converts the relative depth value of a predetermined part of the object to be measured into three-dimensional position coordinates based on the distance calculated by the calculation unit; A control unit that operates an actuator according to the distance obtained by the distance measuring control device and controls the operation, A robot that possesses the following features.

12. A distance measurement control device comprising: an estimation unit that captures images by a monocular camera to include multiple reference points set at different predetermined distances, converts them into a point cloud of predetermined density, and estimates the relative depth value of each point cloud for an object to be measured that exists within the shooting range; a correction unit that corrects the relative depth value estimated by the estimation unit based on the relative depth value of each of the multiple reference points within the point cloud for which the reference relative depth value is known and each of the reference relative depth values; and a conversion unit that converts the relative depth value of a predetermined part of the object to be measured into three-dimensional position coordinates based on the relative depth value corrected by the correction unit. A control unit that operates an actuator according to the distance obtained by the distance measuring control device and controls the operation, A robot that possesses the following features.

13. The robot according to claim 11 or claim 12, which is fitted with a vibration suppression member that suppresses vibration of the plurality of reference points caused by the movement of the robot.

14. The robot according to claim 11 or claim 12, wherein the plurality of reference points are provided outside the operating range of the actuator.

15. The robot according to claim 11 or claim 12, wherein the plurality of reference points are attached to the actuator.

Images