Surveillance camera, camera parameter determination method, and program

Abstract

To accurately improve the image quality to improve the efficiency of monitoring work by adaptively determining the camera parameters that contribute to improving the monitoring accuracy of a human face in an image in an imaging area.SOLUTION: The monitor camera equipped with artificial intelligence, includes: an imaging unit that picks up an imaging area; a detection unit that uses artificial intelligence to detect a person in the captured image in the imaging area; and a control unit that estimates the position of the face of the person in the captured image, and determines the camera parameter of the monitor camera based on the brightness of the face or the movement speed.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present disclosure relates to a surveillance camera that determines camera parameters of a surveillance camera, a method for determining camera parameters, and a program.

Background Art

[0002] Patent Document 1 discloses a mobile remote surveillance camera equipped with artificial intelligence. The mobile remote surveillance camera of Patent Document 1 is an all-in-one structured surveillance camera including a Web camera capable of imaging through a dome-shaped windbreak, a router for mobile communication, a Wi-fi (registered trademark) communication access point device, and an image analysis function using artificial intelligence.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] An area imaged by a surveillance camera (hereinafter referred to as an "imaging area") is often a place that attracts attention for the operator of the surveillance camera (e.g., the owner of a store), such as near an automatic door where people come and go frequently. However, for example, during daytime hours, due to backlighting from sunlight, even if the entire image obtained by imaging the imaging area is bright, the faces of the people shown in the image may be dark, and there is a possibility that the surveillance accuracy deteriorates.

[0005] Furthermore, to improve surveillance accuracy, surveillance cameras that use artificial intelligence (AI) to detect objects being monitored have been proposed (see Patent Document 1). As mentioned above, there was a problem that if the faces of people captured in images by surveillance cameras were dark, the face detection performance by artificial intelligence would decrease, leading to a deterioration in surveillance accuracy. Patent Document 1 does not consider technical measures to improve image quality when the faces of people captured in images are dark. It should be noted that although the above problem was explained using people as an example of subjects, the subjects are not limited to people such as shoplifters, but are thought to be similarly applicable to vehicles that are the target of surveillance. Examples of vehicles that are the target of surveillance include vehicles used by shoplifters to get away.

[0006] This disclosure aims to provide a surveillance camera, a method for determining camera parameters, and a program that, in light of the aforementioned conventional circumstances, adaptively determine camera parameters that contribute to improving the accuracy of monitoring people's faces in the image area, thereby accurately improving image quality and streamlining surveillance operations.

[0007] This disclosure aims to provide a surveillance camera, a method for determining camera parameters, and a program that, in view of the above-mentioned conventional circumstances, adaptively determine camera parameters that contribute to improving the accuracy of monitoring characteristic parts of vehicles captured in images of the imaging area, thereby accurately improving image quality and streamlining surveillance operations. [Means for solving the problem]

[0008] This disclosure relates to a surveillance camera equipped with artificial intelligence, comprising: an imaging unit that captures an image of an imaging area; a detection unit that uses the artificial intelligence to detect a person appearing in the image of the imaging area; and a control unit that estimates the position of the person's face in the image and selectively determines camera parameters of the surveillance camera based on either the brightness of the face or the speed of movement of the face or both, wherein the control unit processes the detection of the presence or absence of the person using the artificial intelligence. The person was detected by this method.The present invention provides a surveillance camera that sets the determined camera parameters to the imaging unit afterward, but before performing the process of detecting whether or not the person's face is present.

[0009] Furthermore, this disclosure relates to a parameter determination method performed by a surveillance camera equipped with artificial intelligence, comprising the steps of: capturing an image of an imaging area; using the artificial intelligence to detect a person appearing in the image of the imaging area; estimating the position of the person's face in the image and selectively determining the camera parameters of the surveillance camera based on either or both of the brightness and movement speed of the face, and processing the detection of the presence or absence of the person using the artificial intelligence. If the person is detected by this method, of detection The determined camera parameters are set to the surveillance camera afterward, but before the process of detecting whether or not the person's face is detected. Step , It further has, This provides a method for determining camera parameters.

[0010] Furthermore, this disclosure describes a process in which a computer, a surveillance camera equipped with artificial intelligence, performs the steps of: capturing an image of an imaging area; using the artificial intelligence to detect a person appearing in the image of the imaging area; estimating the position of the person's face in the image and selectively determining the camera parameters of the surveillance camera based on either the brightness or the movement speed of the face, or both; and detecting the presence or absence of the person using the artificial intelligence. If the person is detected by this method, of detection The determined camera parameters are set to the surveillance camera afterward, but before the process of detecting whether or not the person's face is detected. To execute the next step, We provide a program for that purpose.

[0011] Furthermore, this disclosure relates to a surveillance camera equipped with artificial intelligence, comprising: an imaging unit that images an imaging area; a detection unit that uses the artificial intelligence to detect vehicles appearing in the image of the imaging area; and a control unit that selectively determines camera parameters of the surveillance camera based on either or both the brightness and / or speed of the vehicle in the image, wherein the control unit processes the detection of the presence or absence of the vehicle using the artificial intelligence. The vehicle was detected by The present invention provides a surveillance camera that sets the determined camera parameters in the imaging unit after the fact, but before the processing to detect the vehicle type or color of the vehicle.

[0012] Furthermore, this disclosure relates to a parameter determination method performed by a surveillance camera equipped with artificial intelligence, comprising the steps of: capturing an image of an imaging area; using the artificial intelligence to detect a vehicle appearing in the image of the imaging area; and selectively determining camera parameters of the surveillance camera based on either or both of the brightness and / or speed of the vehicle in the image, wherein the artificial intelligence is used to detect the presence or absence of the vehicle. When the aforementioned vehicle is detected, of detection The determined camera parameters are set to the surveillance camera afterward, but before the process of detecting the vehicle type or color is performed. Step , It further has, This provides a method for determining camera parameters.

[0013] Furthermore, this disclosure describes a process in which a computer, a surveillance camera equipped with artificial intelligence, performs the steps of: capturing an image of an area; using the artificial intelligence to detect a vehicle in the captured image of the area; and selectively determining the camera parameters of the surveillance camera based on either or both the brightness and / or speed of the vehicle in the captured image, thereby detecting the presence or absence of the vehicle using the artificial intelligence. When the aforementioned vehicle is detected, of detection The determined camera parameters are set to the surveillance camera afterward, but before the process of detecting the vehicle type or color is performed. To execute the next step, We provide a program for that purpose.

[0014] Note that these general or specific aspects may be implemented in a system, apparatus, method, integrated circuit, computer program, or recording medium, or in any combination of a system, apparatus, method, integrated circuit, computer program, and recording medium.

Advantages of the Invention

[0015] According to the present disclosure, camera parameters that contribute to improving the monitoring accuracy of the face of a person reflected in an image of an imaging area can be adaptively determined, and the image quality of the image can be accurately improved to improve the efficiency of the monitoring operation.

[0016] According to the present disclosure, camera parameters that contribute to improving the monitoring accuracy of the characteristic parts of a vehicle reflected in an image of an imaging area can be adaptively determined, and the image quality of the image can be accurately improved to improve the efficiency of the monitoring operation.

[0017] Further advantages and effects in one aspect of the present disclosure will be clarified from the specification and drawings. Such advantages and / or effects are provided by several embodiments and the features described in the specification and drawings, respectively, but not necessarily all are provided in order to obtain one or more identical features.

Brief Description of the Drawings

[0018] [Figure 1] A diagram showing an example of the system configuration of a monitoring camera system according to Embodiment 1 [Figure 2] A block diagram showing an example of the hardware configuration of a monitoring camera according to Embodiment 1 [Figure 3] A diagram showing a first proportionality of captured images before and after adjustment of camera parameters when a person is exemplified as a subject [Figure 4] A diagram showing a second proportionality of captured images before and after adjustment of camera parameters when a person is exemplified as a subject [Figure 5] A flowchart showing a first operation procedure example of a monitoring camera according to Embodiment 1 in time series [Figure 6] This diagram shows the proportional relationship between captured images before and after adjusting camera parameters, using a vehicle as an example subject. [Figure 7] A flowchart showing a second example of the operation procedure of the surveillance camera according to Embodiment 1, in chronological order. [Modes for carrying out the invention]

[0019] The following describes in detail embodiments of the surveillance camera, camera parameter determination method, and program disclosed herein, with reference to the drawings as appropriate. However, unnecessary details may be omitted. For example, detailed explanations of already well-known matters or redundant explanations of substantially identical configurations may be omitted. This is to avoid the following explanation becoming unnecessarily verbose and to facilitate understanding by those skilled in the art. The accompanying drawings and the following explanation are provided to enable those skilled in the art to fully understand this disclosure and are not intended to limit the subject matter described in the claims.

[0020] Figure 1 shows an example of the system configuration of a surveillance camera system 100 according to Embodiment 1. As shown in Figure 1, the surveillance camera system 100 includes a surveillance camera 1 and a matching server 50 to which the monitor MN1 and the operation unit MH1 are connected. The surveillance camera 1 and the matching server 50 are connected via a network NW1 so that they can send and receive data from each other.

[0021] Network NW1 is a wireless network conforming to, for example, Wi-Fi® (Registered Trademark), Bluetooth® (Registered Trademark), and WiGig (Wireless Gigabit), but is not limited to these. Network NW1 may also be a wired network such as a USB (Universal Serial Bus) cable or a wired LAN. In the following description, the image captured by surveillance camera 1 (hereinafter referred to as "captured image") includes not only the captured image data but also the camera ID (Identification) of the surveillance camera 1 that captured the image and information on the date and time of capture.

[0022] Surveillance camera 1 is installed in various locations, such as on poles (not shown) on the roadside, in parking lots, and on the ceiling near automatic doors inside buildings such as shops. Surveillance camera 1 captures images of subjects entering the imaging area within a predetermined width of field of view. Surveillance camera 1 is also equipped with artificial intelligence (AI), which is used to detect objects from the captured images of subjects. Objects include, for example, people, people's faces, vehicles, and distinctive parts of vehicles. Distinctive parts of vehicles include, for example, the car model, car color, license plate number, and car manufacturer's emblem. Car model refers to the body type of the vehicle, such as minivan, sedan, or van. Objects are not limited to the people, people's faces, vehicles, and distinctive parts of vehicles mentioned above, but also include a broader range of moving objects such as motorcycles, bicycles, tricycles, and kick scooters, as well as distinctive parts of those moving objects.

[0023] The surveillance camera 1 stores a learning model in the learning model memory 15b (see Figure 2). This learning model is a set of parameters that characterize the functions of the AI ​​installed in the surveillance camera 1. In other words, the learning model is a set of parameters that determine the objects that the AI ​​installed in the surveillance camera 1 detects. The learning model may also be, for example, a set of parameters that determine the structure of the neural network N1 of the surveillance camera 1. The set of parameters that determine the structure of the neural network N1 of the surveillance camera 1 may include, for example, information indicating the connection relationships between units of the neural network N1 or weight coefficients. The learning model is generated by a terminal device (not shown) and stored in the surveillance camera 1 beforehand.

[0024] The surveillance camera 1 can detect objects by constructing a corresponding neural network N1 for each object to be detected. For example, the surveillance camera 1 can perform object detection by constructing a neural network for detecting people, a neural network for detecting people's faces, a neural network for detecting vehicles, and a neural network for detecting characteristic parts of vehicles. The neural network N1 shown in Figure 1 is a collective term for the four individual neural networks exemplified above.

[0025] When surveillance camera 1 detects, for example, a person's face, it generates a face-extracted image by cutting out the face from the captured image and sends a matching instruction including the face-extracted image to the matching server 50 via the network NW1.

[0026] When surveillance camera 1 recognizes the characters of a vehicle's license plate, for example, it sends a matching instruction, including the character recognition result of the license plate, to the matching server 50 via the network NW1.

[0027] The matching server 50 is an information processing device such as a personal computer, smartphone, tablet terminal, or a server computer machine with high specifications. The matching server 50 communicates data with the surveillance camera 1 via the network NW1.

[0028] The matching server 50 includes a communication IF circuit 51, a processor 52, memory 53, and a database 54. Note that in the attached diagram, the interface is conveniently abbreviated as "IF".

[0029] The communication IF circuit 51 communicates data with the surveillance camera 1 via the network NW1 described above. The communication IF circuit 51 receives, for example, a verification instruction (see above) sent from the surveillance camera 1 and outputs it to the processor 52.

[0030] The processor 52 is configured using, for example, a CPU (Central Processing Unit), a DSP (Digital Signal Processor), a GPU (Graphical Processing Unit), or an FPGA (Field Programmable Gate Array). The processor 52 functions as a controller that oversees the overall operation of the matching server 50, performing control processing to coordinate the operation of each part of the matching server 50, data input / output processing between each part of the matching server 50, data calculation processing, and data storage processing. The processor 52 operates according to the programs and data stored in the memory 53. The processor 52 uses the memory 53 during operation and temporarily stores data or information that the processor 52 generates or acquires in the memory 53.

[0031] Based on the matching instruction from the communication IF circuit 51, the processor 52 performs a process (i.e., a matching process) to determine whether a face-removed image or a number character recognition result included in the matching instruction is registered in the database 54. The processor 52 may also perform this matching process using an AI function that utilizes a learning model created for the matching process. This allows the processor 52 to perform a highly accurate matching process. The processor 52 may also respond to the results of the matching process via the network NW1 to the surveillance camera 1 or other external device (for example, a terminal held by the person requesting the matching process), or it may output (display) the results on the monitor MN1. The matching of face-removed images sent from the surveillance camera 1 may be performed by the matching process of the matching server 50, or it may be performed by a person (for example, a system administrator) using a receiving terminal (not shown) that receives the face-removed images sent from the surveillance camera 1 to visually confirm them.

[0032] Memory 53 is configured, for example, using RAM (Random Access Memory) and ROM (Read Only Memory), and temporarily stores programs necessary for the operation of the matching server 50, as well as data or information generated during operation. RAM is, for example, work memory used when the matching server 50 is operating. ROM stores, for example, programs for controlling the matching server 50 in advance.

[0033] Database 54 is configured using, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive). Database 54 stores face blacklist data (not shown) containing blacklists of faces of individuals to be matched, or vehicle license plate blacklists (not shown) containing blacklists of license plate numbers of vehicles to be matched.

[0034] Monitor MN1 is a display device configured using, for example, an LCD (Liquid Crystal Display) or an organic EL (Electroluminescence). Monitor MN1 displays, for example, the results of the matching process performed by the matching server 50. Note that Monitor MN1 may be included in the matching server 50.

[0035] The operation unit MH1 is an input device that accepts user input operations, such as a mouse, keyboard, touchpad, or touch panel. The operation unit MH1 sends signals corresponding to the user input operations to the matching server 50. The operation unit MH1 may also be included in the matching server 50.

[0036] Figure 2 is a block diagram showing an example of the hardware configuration of the surveillance camera 1 according to Embodiment 1. In addition to the surveillance camera 1, Figure 2 also shows an external storage medium M1 that is inserted into and removed from the surveillance camera 1. The external storage medium M1 is a storage medium such as an SD card.

[0037] As shown in Figure 2, the surveillance camera 1 includes a lens 11, a lens drive unit LM1, an image sensor 12, a memory 13, a processor 14, an AI processing unit 15, a communication IF circuit 16, an external storage medium IF 18, an IR illumination unit 19, and a pan-tilt drive unit PTM1.

[0038] An example of an imaging unit, the lens 11, includes, for example, a focus lens and a zoom lens, and receives incident light ICL1, which is light reflected by the subject, to form an optical image of the subject on the light-receiving surface of the image sensor 12. Depending on the installation location or shooting application of the surveillance camera 1, lenses with various focal lengths or shooting ranges can be used for the lens 11.

[0039] As an example of an imaging unit, the lens drive unit LM1 drives the lens 11 to adjust camera parameters related to the lens 11 (e.g., the position of the focus lens, the position of the zoom lens corresponding to the zoom magnification) based on control signals from the processor 14.

[0040] An image sensor 12, as an example of an imaging unit, sets camera parameters related to the image sensor 12 (e.g., exposure time corresponding to shutter speed, gain) and uses these camera parameters to perform photoelectric conversion to convert light received on the light-receiving surface into an electrical signal. The image sensor 12 is, for example, an image sensor such as a CCD (Charge Coupled Device) or CMOS (Complementary Metal Oxide Semiconductor). The image sensor 12 outputs an electrical signal (analog signal) corresponding to the light received on the light-receiving surface to the processor 14. This analog signal is converted into digital image data by the processor 14, as an example of an imaging unit. As a result, the image data is generated by the processor 14.

[0041] Memory 13 is configured, for example, using RAM and ROM, and temporarily stores programs necessary for the operation of the surveillance camera 1, as well as data or information generated during operation. RAM is, for example, work memory used when the surveillance camera 1 is operating. ROM stores, for example, the program according to this disclosure for controlling the surveillance camera 1 in advance. In other words, by executing the program stored in ROM, the processor 14 can enable the surveillance camera 1, which is a computer, to perform various processes according to the camera parameter determination method according to this disclosure.

[0042] The processor 14 is configured using, for example, a CPU, DSP, GPU, or FPGA. The processor 14 functions as a controller that oversees the overall operation of the surveillance camera 1, performing control processing to coordinate the operation of each part of the surveillance camera 1, data input / output processing between each part of the surveillance camera 1, data calculation processing, and data storage processing. The processor 14 operates according to the programs and data stored in the memory 13. The processor 14 uses the memory 13 during operation and temporarily stores data or information that the processor 14 generates or acquires in the memory 13.

[0043] Furthermore, the processor 14, as an example of an imaging unit, generates digital image data by applying predetermined signal processing to the electrical signal output from the image sensor 12 and outputs it to the AI ​​processing unit 15. At this time, the processor 14 may generate the image data by adjusting (changing) camera parameters related to signal processing (e.g., tone curve) before executing the signal processing.

[0044] Furthermore, the processor 14 has a timer (not shown) and can determine the current time based on the timer's output. During the night (in other words, from around sunset to around dawn), it outputs a control signal to the IR illumination unit 19 instructing it to irradiate with IR light.

[0045] Furthermore, the processor 14, as an example of a control unit, decides to adjust (change) the camera parameters of the surveillance camera 1 based on the processing results of the AI ​​processing unit 15 (details are described later). The processor 14 generates a control signal to adjust (change) the determined camera parameters and outputs it to at least one of the lens drive unit LM1, image sensor 12, signal processing unit within the processor 14, and pan-tilt drive unit PTM1. For example, if the AI ​​processing unit 15 cannot detect an object (e.g., a person's face) that should be visible in the captured image due to the effects of backlighting, the processor 14 adjusts (changes) the current camera parameters so that the AI ​​processing unit 15 can improve the detection accuracy of that object (e.g., a person's face).

[0046] Here, the camera parameters that are adjusted (changed) include, for example, camera parameters related to the lens 11 (see above), camera parameters related to the image sensor 12 (see above), camera parameters related to signal processing, camera parameters related to pan rotation or tilt rotation, and camera parameters related to the IR illumination unit 19. However, it goes without saying that the camera parameters that are adjusted (changed) are not limited to the camera parameters mentioned above. Details of the adjustment (change) of camera parameters by the processor 14 based on the processing results of the AI ​​processing unit 15 will be described later.

[0047] As an example of a detection unit, the AI ​​processing unit 15 uses artificial intelligence to detect objects to be detected from the image data output from the processor 14. The AI ​​processing unit 15 is composed of a processor such as a CPU, DSP, GPU, or FPGA, and memory such as flash memory. The AI ​​processing unit 15 has an AI calculation processing unit 15a and a learning model memory 15b.

[0048] The AI ​​processing unit 15a forms an AI (artificial intelligence) based on the learning model stored in the learning model memory 15b. For example, the AI ​​processing unit 15a forms a neural network N1 (see Figure 1) based on the learning model. The AI ​​processing unit 15a receives the image data of the subject generated by the processor 14, and the neural network N1 based on the learning model detects the object to be detected from the input image data.

[0049] The learning model memory 15b is composed of memory such as RAM, ROM, and flash memory. The learning model memory 15b stores learning models for object detection processing that have been created in advance through a learning process. The AI ​​processing unit 15a can detect objects in the captured image by executing the learning model for each object from the learning model memory 15b to form the corresponding neural network N1.

[0050] As an example of a communication unit, the communication IF circuit 16 performs data communication (sending and receiving) with the matching server 50 connected via the network NW1. The communication IF circuit 16 sends matching instructions (for example, requests for matching a face-extracted image with face blacklist data, or matching the character recognition result of a number with number blacklist data) generated by the processor 14 to the matching server 50.

[0051] External storage media such as an SD card (M1) can be inserted into or removed from the external storage media IF18.

[0052] The IR illumination unit 19 starts irradiating the imaging area with IR light RD1 having a near-infrared wavelength band based on a control signal from the processor 14 (for example, an instruction to start IR light irradiation). The IR illumination unit 19 stops irradiating the imaging area with IR light RD1 based on a control signal from the processor 14 (for example, an instruction to end the irradiation of IR light RD1). The IR illumination unit 19 also increases or decreases the intensity of the currently irradiated IR light RD1 based on a control signal from the processor 14 (for example, an instruction to adjust the intensity of IR light RD1).

[0053] The pan-tilt drive unit PTM1 adjusts camera parameters related to pan rotation or tilt rotation (e.g., pan rotation amount, tilt rotation amount) based on control signals from the processor 14 to perform pan rotation, tilt rotation, or both of the surveillance camera 1.

[0054] Figure 3 shows the first proportional relationship between captured images before and after adjusting camera parameters, with a person as an example subject. Captured image IMG1 is, for example, an image captured by surveillance camera 1 with a person PS1 inside a building as the subject. Surveillance camera 1 can detect the entire frame of person PS1 or the frame of person PS1 above the shoulders SD1 (i.e., above the shoulder blades) by AI processing using captured image IMG1, which was captured based on the camera parameters currently set to optimally capture the entire imaging area within the field of view.

[0055] However, under the camera parameters at the time the captured image IMG1 was taken, the face FC1 of person PS1 is captured dark due to the effects of backlighting such as sunlight. In this case, the face FC1 of person PS1 cannot be specifically identified, and the accuracy of face matching decreases. As a result, the monitoring accuracy of the captured area is likely to deteriorate. It should be noted that even if the face FC1 is captured darkly, the surveillance camera 1 can still detect person PS1 in the captured image IMG1 as described above, so it can acquire brightness (luminosity) information of the face FC1 area.

[0056] Therefore, in the surveillance camera 1 according to Embodiment 1, the processor 14 adjusts (changes) the camera parameters (for example, the exposure time, gain, or tone curve characteristics of the image sensor 12) from their current values ​​to increase or sharpen them, using the brightness information of the face FC1 region, so that the face region can be imaged brightly. The tone curve is used in signal processing performed by the processor 14 (for example, generation of image data) and represents a characteristic that defines the brightness gradation of the output relative to the brightness gradation of the input. The tone curve characteristics are stored, for example, in the processor 14 or in memory 13.

[0057] According to the captured image IMG2 taken by the surveillance camera 1 after this adjustment (modification), the area of ​​the face FC2 of the same person PS1 becomes brighter (in other words, the brightness of face FC2 increases), and a brightness suitable for face matching (face recognition) processing at the matching server 50 is obtained. In other words, the AI ​​processing unit 15 can detect face FC2 with a high probability. Therefore, it is expected that the accuracy of face matching performed at the matching server 50 will also improve. Furthermore, according to the captured image IMG2 described above, the surveillance camera 1 can improve the detection (identification) accuracy of the whole body or shoulder-up SD2 of person PS1 captured in the captured image IMG2 (for example, the detection accuracy of details such as gender, age, facial features, and clothing, not just the face). Note that even if the brightness of face FC2 increases, the AI ​​processing unit 15 can still detect the entire frame of person PS1 or the frame of shoulder-up SD2 of person PS1 (i.e., above the shoulder blades) in the same way as in the captured image IMG1.

[0058] Figure 4 shows the second proportional relationship between captured images before and after adjusting camera parameters, with a person as an example subject. Captured image IMG3 is an image captured by surveillance camera 1 with a person PS2 riding a bicycle outdoors as the subject. Surveillance camera 1 captures images based on the currently set camera parameters (e.g., exposure time 1 / 30 [seconds]) in order to optimally capture the entire imaging area within the field of view. As a result, as shown in captured image IMG3, the image of person PS2 moving within the field of view is blurred. However, if the degree of blur is slight, the entire frame of person PS2 can be detected by AI processing. On the other hand, in blurred images, it is difficult to detect the face of person PS2.

[0059] Therefore, in the surveillance camera 1 according to Embodiment 1, the processor 14 adjusts (changes) the camera parameters (e.g., exposure time of the image sensor 12) from the current value (e.g., 1 / 30 [seconds]) to another set value (e.g., 1 / 640 [seconds]) using the movement speed and direction of the person PS2 so that the person PS2 can be imaged without blurring. The adjusted set value is stored, for example, in the processor 14 or in the memory 13.

[0060] According to the image IMG4 captured by the surveillance camera 1 after this adjustment (modification), the same person PS2 can be captured clearly without blurring, resulting in a clearer image of the person PS2's face, and providing a blur-free image suitable for face matching (face recognition) processing on the matching server 50. In other words, the AI ​​processing unit 15 can detect the face of person PS2 with a high probability. Even if the blurring of the person PS2's face is suppressed, the AI ​​processing unit 15 can still detect the entire frame of person PS2, similar to the image IMG3.

[0061] Next, a first example of the operation procedure of the surveillance camera 1 according to Embodiment 1 will be described with reference to Figure 4. Figure 5 is a flowchart showing the first example of the operation procedure of the surveillance camera 1 according to Embodiment 1 in chronological order. In the first example of the operation procedure, the subject of the surveillance camera 1 is a person, and an example is described in which the camera parameters are adjusted (changed) so that the person's face can be detected by AI processing. In the explanation of Figure 5, the camera parameters to be adjusted (changed) are exemplified as the exposure time and gain and tone curve of the image sensor 12, but it goes without saying that the parameters are not limited to these.

[0062] In Figure 5, the surveillance camera 1 sets the exposure time and gain of the image sensor 12 to predetermined values ​​(for example, default initial values ​​that enable the entire captured image to be brightly captured) as an example of setting camera parameters (St1). The surveillance camera 1 also sets a tone curve with default characteristics (for example, default characteristics that enable the entire captured image to be brightly captured) as an example of camera parameters (St2). After setting the camera parameters in steps St1 and St2, the surveillance camera 1 photographs (images) the subject (for example, a person PS1).

[0063] The surveillance camera 1 performs AI-based detection processing on the captured image data of the subject (e.g., person PS1) (St3). The surveillance camera 1 determines whether or not it has detected person PS1 (see Figure 3, e.g., the entire image of person PS1) in the captured image through the AI-based detection processing (St4). Note that this entire image of person PS1 may be an entire image identified according to the field of view from the mounting position of the surveillance camera 1, and is not limited to an entire image viewed from the front of person PS1, but may also be an entire image viewed from the side, diagonally in front, or diagonally behind.

[0064] If the surveillance camera 1 detects a person PS1 (see Figure 3) in the captured image (St4, YES), it estimates the position of the person PS1's face from the detection frame of the person PS1 identified by detection (i.e., a frame showing the entire body or the area above the shoulders of the detected person PS1) (St5). Furthermore, the surveillance camera 1 identifies the brightness and movement speed of the face in the captured image from the estimated face position, and calculates and determines to increase or decrease the setting values ​​of the image sensor 12 (e.g., exposure time, gain) based on the values ​​of the face brightness and movement speed (St5). Note that the determination in step St5 may be made using only one of the face brightness or movement speed, for example. The surveillance camera 1 may immediately reflect this determined setting value (an example of camera parameters) in the image sensor 12 (St6). This step St6 is optional and may be omitted. If omitted, the setting value determined in step St5 will be reflected to be set in the image sensor 12 at the time of step St1 of step St7 and subsequent steps described later.

[0065] On the other hand, after step St5 (or step St6) or if person PS1 is not detected (St4, NO), the surveillance camera 1 determines whether or not it has detected the face FC2 of person PS1 (see Figure 3) in the captured image through AI detection processing (St7).

[0066] If surveillance camera 1 detects the face FC2 of person PS1 (see Figure 3) in the captured image (St7, YES), it determines the characteristics of the tone curve to be sharper than the current characteristics so that the area of ​​face FC2 in the captured image is optimized (for example, so that it has a default brightness suitable for face matching) (St8). In other words, even if face FC2 is detected by AI in the captured image, if the brightness of the area of ​​face FC2 does not have a default brightness suitable for matching, the processing in step St9, described later, acts as a filtering function, and the tone curve is adjusted (changed) so that the brightness of the area of ​​face FC2 has a default brightness suitable for matching.

[0067] Furthermore, after step St8, the surveillance camera 1 determines whether the face score (for example, a probability output value that quantitatively indicates the likelihood of a face being a face in the face region detected in the captured image) related to the detection of face FC2 by the AI ​​detection process is above a threshold (St9). The threshold is, for example, the probability output value when a face is detected in an captured image that has sufficient brightness to be suitable for face matching processing.

[0068] If surveillance camera 1 determines that the face score when face FC2 is detected is above a threshold (St9, YES), it generates a face cropped image by cutting out the area of ​​face FC2 from the captured image. Surveillance camera 1 generates a matching instruction containing the data of this face cropped image and transfers it to the matching server 50 (St10). After this, the matching server 50 uses AI-based face matching to compare the data of the face cropped image included in the matching instruction transferred from surveillance camera 1 with the face blacklist data. If face blacklist data is stored in surveillance camera 1, the matching process between the face cropped image and the face blacklist data may be performed by surveillance camera 1.

[0069] On the other hand, if face FC2 cannot be detected (St7, NO), or if the face score after step St10 or when face FC2 is detected is determined to be below the threshold (St9, NO), the processing of surveillance camera 1 returns to step St1, and thereafter, the processing of steps St1 to St10 is repeated as a loop. In Figure 5, the exposure time, gain, and tone curve of the image sensor 12 are shown as examples of camera parameters, but the system is not limited to these. For example, pan rotation, tilt rotation, or both of the surveillance camera 1 may be performed, or the position of the focus lens (i.e., aperture control), the position of the zoom lens (i.e., zoom magnification control), or the intensity of the IR light from the IR illumination unit 19 may be adjusted (on, off, dimming).

[0070] As described above, the surveillance camera 1 according to Embodiment 1 is equipped with artificial intelligence and includes an imaging unit that captures an image of the imaging area, a detection unit that uses artificial intelligence to detect a person PS1 in the image of the imaging area, and a unit that estimates the position of the person's face in the image and determines the camera parameters of the surveillance camera 1 based on the brightness or movement speed of the face.

[0071] As a result, surveillance camera 1 can adaptively determine camera parameters that contribute to improving the monitoring accuracy of the imaging area, even in environments where the face of a person PS1 in the image of the imaging area is darkened due to the effects of, for example, backlighting from sunlight, or in scenes with movement such as when a person is riding a moving object such as a bicycle or motorcycle. Therefore, it is possible to improve the detection accuracy of objects targeted for detection in the captured image by artificial intelligence (AI). Consequently, surveillance camera 1 can accurately improve the image quality of the captured image of person PS1 and improve the efficiency of monitoring operations within the imaging area, such as detecting the face of person PS1.

[0072] Furthermore, surveillance camera 1 sets the determined camera parameters. This allows surveillance camera 1 to quickly improve the AI ​​detection accuracy of the object to be detected (for example, the face of person PS1), thereby enabling high-precision detection of the object to be detected (for example, the face of person PS1) in the captured image.

[0073] Furthermore, if the surveillance camera 1 cannot detect a person PS1 in the captured image, it determines and sets the camera parameters so that the person PS1 captured in the image by the imaging unit can be detected by the AI. In this way, if the surveillance camera 1 is installed in an environment where the currently set camera parameters cannot be detected by the AI, it can adjust (change) the camera parameters so that at least the person PS1 can be detected.

[0074] Furthermore, the camera parameter is the exposure time of the imaging unit. The surveillance camera 1 determines the exposure time so that the face of person PS1 in the captured image captured by the imaging unit can be detected by AI (for example, so that it has brightness and blur of a degree suitable for face matching). As a result, the surveillance camera 1 can not only improve the accuracy of AI face detection by making the captured image obtained from the image sensor 12 clearer and accurately improving the image quality of characteristic parts of the subject (for example, the face), but it can also generate face-extracted images with a brightness or blur of a degree suitable for matching using face blacklist data on the matching server 50 with high accuracy.

[0075] Furthermore, the camera parameter is the gain of the imaging unit. The surveillance camera 1 determines the gain so that the faces of people captured in the images captured by the imaging unit can be detected by AI (for example, so that the brightness and degree of blur are suitable for face matching). As a result, the brightness of the images obtained from the image sensor 12 of the surveillance camera 1 is relatively increased, and the image quality of characteristic parts of the subject (for example, faces) is accurately improved. This not only improves the accuracy of AI-based face detection but also enables the generation of highly accurate face-extracted images with sufficient brightness or degree of blur to be used in the matching process using face blacklist data on the matching server 50.

[0076] Furthermore, the camera parameter is the tone curve used to generate the captured image. The surveillance camera 1 determines the tone curve so that the face of person PS1 in the captured image captured by the imaging unit can be detected by AI (for example, so that it has brightness and blur of a degree suitable for face matching). As a result, the brightness of the captured image generated by the processor 14 is relatively increased, and the image quality of characteristic parts of the subject (for example, the face) is accurately improved. This not only improves the accuracy of AI-based face detection but also enables the generation of highly accurate face-extracted images with brightness or blur of a degree suitable for matching using face blacklist data on the matching server 50.

[0077] Furthermore, the camera parameters are at least one of the pan rotation and tilt rotation of the imaging unit. The surveillance camera 1 determines the amount of rotation so that the face of person PS1 in the image captured by the imaging unit can be detected by the AI. As a result, the surveillance camera 1 can properly include the person's face within the field of view, thereby accurately improving the accuracy of AI-based face detection. In addition, the surveillance camera 1 can generate face-extracted images with high accuracy that have a resolution sufficient for use in matching processing using face blacklist data on the matching server 50.

[0078] Furthermore, the camera parameter is the zoom magnification of the imaging unit. Surveillance camera 1 determines the zoom magnification so that the face of person PS1 in the image captured by the imaging unit can be detected by AI. As a result, even if the proportion of person PS1's face in the captured image is small, surveillance camera 1 can detect the person's face with high accuracy by increasing the zoom magnification and zooming in. In addition, surveillance camera 1 can generate face-extracted images with high accuracy that have a resolution sufficient for matching processing using face blacklist data on the matching server 50.

[0079] Furthermore, if, after the camera parameters are set, the surveillance camera 1 detects a person's face in the image captured by the imaging unit using AI and the face detection score (e.g., face score) is above a threshold, it sends a face-extracted image, obtained by cutting out the face from the image, to a face matching server (e.g., matching server 50) that is communicably connected to the surveillance camera 1. This allows the surveillance camera 1 to generate a face-extracted image of a person who has entered the imaging area, and to request the matching server 50 to determine whether the face in this face-extracted image is the face of a suspicious person registered in the face blacklist data. Consequently, the monitoring accuracy of the imaging area is effectively improved.

[0080] Figure 6 shows the proportional relationship between captured images before and after adjusting camera parameters, with a vehicle as an example subject. Captured image IMG5 is an image taken by a surveillance camera 1 installed on the side of a road at night, with a vehicle VCL1 entering the imaging area as the subject. The surveillance camera 1 can detect the silhouette of the vehicle VCL1 (i.e., a blurry overall frame) by AI processing using captured image IMG5, which was captured based on the camera parameters currently set to optimally capture the entire imaging area within the field of view.

[0081] However, under the camera parameters at the time the captured image IMG5 was taken, it is possible that the rear license plate and its surroundings of vehicle VCL1 may be overexposed due to the influence of, for example, following vehicles at night, oncoming vehicles, or the headlights of the own vehicle, or that vehicle VCL1 may be photographed blurred while in motion, making it difficult for the AI ​​to detect not only the license plate but also the details of vehicle VCL1. In such cases, even if the human eye can recognize the car's shape and color, the AI ​​processing within the surveillance camera 1 may not be able to detect the vehicle VCL1's unique features such as its license plate, shape, and color with high accuracy, and the monitoring accuracy of the imaged area is likely to deteriorate.

[0082] Therefore, in the surveillance camera 1 according to Embodiment 1, the processor 14 first adjusts (changes) the camera parameters (for example, the exposure time, gain, and tone curve characteristics of the image sensor 12) from their current values ​​to increase or sharpen them, so that the overall external feature parts of the vehicle VCL1 can be captured with high accuracy, prioritizing them over more detailed partial feature parts such as license plates. These adjusted camera parameters are set in the image sensor 12 or stored in the processor 14 or memory 13.

[0083] According to the image IMG6 captured by the surveillance camera 1 after this adjustment (modification), the area of ​​the entire exterior (outer shape) of the vehicle VCL1 becomes brighter. As a characteristic feature of the vehicle VCL1's exterior, the surveillance camera 1 can determine the vehicle type (e.g., minivan) and color (e.g., pink) of the vehicle VCL1 from the image IMG6 through AI processing. In other words, the AI ​​processing unit 15 can detect the vehicle type and color of the vehicle VCL1 with high probability.

[0084] Next, the processor 14 adjusts (changes) the camera parameters (for example, the position of the zoom lens corresponding to the zoom magnification) from their current value to zoom up, so that partial feature parts of the vehicle VCL1, such as the license plate or emblem (in this case, the license plate), can be captured with high precision. These adjusted camera parameters are stored in the lens drive unit LM1, the processor 14, or the memory 13. As a result, the surveillance camera 1 can perform a zoom-up process so that the license plate of the vehicle VCL1 is in the center of the captured image compared to when the captured image IMG6 was taken.

[0085] According to the image IMG7 captured by the surveillance camera 1 after this zoom-up process, the area of ​​the license plate of the same vehicle VCL1 becomes brighter and more enlarged. As a partial feature part of vehicle VCL1, the surveillance camera 1 can identify the license plate of vehicle VCL1 (for example, Shinagawa 306, ho, 14-71) from the AI-captured image IMG7. In other words, the AI ​​processing unit 15 can detect the license plate of vehicle VCL1 with increasing probability in stages, following the vehicle type and color of vehicle VCL1.

[0086] Next, a first example of the operation procedure of the surveillance camera 1 according to Embodiment 1 will be described with reference to Figure 7. Figure 7 is a flowchart showing a second example of the operation procedure of the surveillance camera 1 according to Embodiment 1 in chronological order. In the second example of the operation procedure, the subject of the surveillance camera 1 is a vehicle, and an example is described in which the camera parameters are adjusted (changed) so that characteristic parts of the vehicle (for example, the car model, car color, and license plate in that order) can be detected in stages by AI processing. In the explanation of Figure 7, the camera parameters to be adjusted (changed) are exemplified as the exposure time and gain of the image sensor 12, the tone curve, the pan rotation amount, the tilt rotation amount, and the position of the zoom lens (in other words, the zoom magnification), but it goes without saying that this is not limited to these. Also, in the explanation of Figure 7, the explanation of processes that overlap with the explanation of Figure 4 is given the same step number to simplify or omit the explanation, and different content is explained.

[0087] In Figure 7, after steps St1 and St2, the surveillance camera 1 performs PTZ (Pan Tilt Zoom) control (St11) to reflect the setting of at least one of the following camera parameters: pan rotation amount, tilt rotation amount, and zoom magnification. After setting the camera parameters in steps St1, St2, and St11, the surveillance camera 1 photographs (images) the subject (e.g., vehicle VCL1). Note that the processing in step St11 is optional for the surveillance camera 1, and it is not always necessary to perform PTZ control; it may be performed only when there is a change in the camera parameters related to PTZ control.

[0088] The surveillance camera 1 performs AI-based detection processing on the image data of the subject (e.g., vehicle VCL1) (St3). The surveillance camera 1 determines whether or not it has detected vehicle VCL1 (see Figure 6, e.g., the entire view of vehicle VCL1 or a part of vehicle VLC1) in the captured image through AI-based detection processing (St12). The entire view of vehicle VCL1 may be the entire view or a part of it, as determined by the angle of view from the mounting position of the surveillance camera 1, and is not limited to the entire view of vehicle VCL1 from the front, but may also be the entire view or a part of it viewed from the side, diagonally in front, or diagonally in rear.

[0089] If the surveillance camera 1 detects a vehicle VCL1 (see Figure 6) in the captured image (St12, YES), it calculates and determines whether to increase or decrease the setting value of the image sensor 12 (e.g., exposure time, gain) based on whether the vehicle VCL1's headlights are on in the captured image, the brightness of the headlight area, or the vehicle VCL1's movement speed (St13). The surveillance camera 1 may immediately reflect this determined setting value (an example of camera parameters) in the image sensor 12 (St6). This step St6 is optional and may be omitted. If omitted, the setting value determined in step St12 will be reflected in the image sensor 12 at the time of step St14 and subsequent steps described later.

[0090] On the other hand, after step St13 (or step St6) or if vehicle VCL1 cannot be detected (St12, NO), the surveillance camera 1 uses AI detection processing to determine whether or not the vehicle type and color of vehicle VCL1 (see Figure 6) could be identified in the captured image (St14). Note that in step St14, it is determined whether or not the vehicle type and color have been identified, but it is also acceptable to determine whether or not only the vehicle type or color has been identified.

[0091] If surveillance camera 1 identifies the vehicle type and color of vehicle VCL1 (see Figure 6) in the captured image (St14, YES), it determines whether or not it was able to detect the license plate through AI-based detection processing (identification processing) (St15).

[0092] If surveillance camera 1 detects the license plate of vehicle VCL1 (St15, YES), it generates a license plate extract image by cutting out the license plate portion from the captured image. Surveillance camera 1 generates a matching instruction including the license plate extract image and transfers it to the matching server 50 (St16). After this, the matching server 50 performs character recognition processing using AI and compares the data of the license plate extract image with the license plate blacklist data. Note that at least one of the character recognition processing and matching processing may be performed by surveillance camera 1 (for example, processor 14 as an example of a character recognition unit and a matching unit).

[0093] On the other hand, if the license plate of vehicle VCL1 cannot be detected (St15, NO), the surveillance camera 1 calculates the zoom magnification, pan rotation amount, and tilt rotation amount of the lens based on the size of the detected vehicle VCL1 so that the size of the license plate in the screen is suitable for license plate recognition (St17). The surveillance camera 1 may immediately reflect these determined settings in the PTZ control (St18). This step St18 is optional and may be omitted. If omitted, the settings determined in step St17 will be reflected in the lens drive unit LM1 and pan-tilt drive unit PTM1 at the time of the subsequent step St11. This zoom-up process for the license plate area is expected to improve the probability of detecting license plates in frames of captured images that are captured and input in subsequent steps.

[0094] After step St16, after step St17 (or step St18), or if it is determined that neither the vehicle type nor the vehicle color was detected (St14, NO), the processing of the surveillance camera 1 returns to step St1, and thereafter, the processing of steps St1 to St18 is repeated as a loop. In Figure 7, the exposure time, gain, tone curve, pan rotation amount, tilt rotation amount, and zoom lens position of the image sensor 12 are shown as examples of camera parameters, but the system is not limited to these, and for example, the position of the focus lens (i.e., aperture control) or the intensity of the IR light from the IR illumination unit 19 can be adjusted (on, off, dimming).

[0095] As described above, the surveillance camera 1 according to Embodiment 1 is equipped with artificial intelligence and includes an imaging unit that captures an image of the imaging area, a detection unit that uses artificial intelligence to detect the vehicle VCL1 that appears in the image of the imaging area, and determines the camera parameters of the surveillance camera 1 based on the brightness of the vehicle VCL1 in the image (for example, whether the headlights are on or off, and the brightness of the headlight area when the headlights are on) or the speed of movement of the vehicle VCL1 in the image.

[0096] As a result, surveillance camera 1 can adaptively determine camera parameters that contribute to improving the monitoring accuracy of the imaging area, even in environments where characteristic parts of the vehicle VCL1 appear in the image of the imaging area are darkened due to the influence of, for example, following vehicles at night, oncoming vehicles, or the headlights of its own vehicle, or in environments where blurring occurs due to the movement of the vehicle on the screen of the image. Therefore, it is possible to improve the detection accuracy of objects to be detected in the image by artificial intelligence (AI). Consequently, surveillance camera 1 can accurately improve the image quality of the captured vehicle VCL1 image and improve the efficiency of monitoring operations within the imaging area, such as detecting at least the vehicle type and color of the vehicle VCL1.

[0097] Furthermore, surveillance camera 1 sets the determined camera parameters. This allows surveillance camera 1 to quickly improve the AI ​​detection accuracy of the objects to be detected (e.g., the vehicle type and color of vehicle VCL1, and the license plate), enabling high-precision detection of the objects to be detected in the captured image (e.g., the vehicle type and color of vehicle VCL1, and the license plate).

[0098] Furthermore, if the surveillance camera 1 cannot detect the vehicle VCL1 in the captured image, it determines and sets the camera parameters so that the vehicle VCL1 captured in the image taken by the imaging unit can be detected by the AI. In this way, if the surveillance camera 1 is installed in an environment where the currently set camera parameters cannot be detected by the AI, it can adjust (change) the camera parameters so that at least the vehicle VCL1 can be detected.

[0099] Furthermore, the camera parameter is the exposure time of the imaging unit. The surveillance camera 1 determines the exposure time so that the vehicle type, color, make, or license plate of the vehicle VCL1 captured in the image taken by the imaging unit can be detected by AI. As a result, the image obtained from the image sensor 12 of the surveillance camera 1 becomes clearer, and the image quality of characteristic parts of the subject (e.g., vehicle type, color, make, license plate) is accurately improved. This not only improves the accuracy of AI detection of characteristic parts of the vehicle, but also enables the generation of license plate cutout images with sufficient brightness to be used in matching processing using license plate blacklist data on the matching server 50 with high accuracy.

[0100] Furthermore, the camera parameter is the gain of the imaging unit. The surveillance camera 1 determines the gain so that the vehicle type, color, make, or license plate of the vehicle VCL1 captured in the image taken by the imaging unit can be detected by AI. As a result, the brightness of the image obtained from the image sensor 12 of the surveillance camera 1 is relatively increased, and the image quality of characteristic parts of the subject (e.g., vehicle type, color, make, license plate) is accurately improved. This not only improves the accuracy of AI detection of characteristic parts of the vehicle, but also enables the generation of license plate cutout images with sufficient brightness to be used in matching processing using license plate blacklist data on the matching server 50 with high accuracy.

[0101] Furthermore, the camera parameter is the tone curve used to generate the captured image. The surveillance camera 1 determines the tone curve so that the vehicle type, color, make, or license plate of the vehicle VCL1 captured in the image taken by the imaging unit can be detected by AI. As a result, the brightness of the captured image generated by the processor 14 is relatively increased, and the image quality of characteristic parts of the subject (e.g., vehicle type, color, make, license plate) is accurately improved. This not only improves the accuracy of AI detection of characteristic parts of the vehicle, but also enables the generation of license plate cutout images with sufficient brightness for use in matching processing using license plate blacklist data on the matching server 50 with high accuracy.

[0102] Furthermore, the camera parameters are at least one of the pan rotation and tilt rotation of the imaging unit. The surveillance camera 1 determines the amount of rotation so that the vehicle type, color, make, or license plate of the vehicle VCL1 captured in the image taken by the imaging unit can be detected by AI. As a result, the surveillance camera 1 can properly include the vehicle within the field of view and accurately improve the accuracy of AI detection of characteristic parts of the vehicle (e.g., vehicle type, color, make, license plate).

[0103] Furthermore, the camera parameter is the zoom magnification of the imaging unit. Surveillance camera 1 determines the zoom magnification so that the vehicle type, color, make, or license plate of vehicle VCL1 captured in the image taken by the imaging unit can be detected by AI. As a result, even if vehicle VCL1 occupies a small proportion of the captured image, surveillance camera 1 can increase the zoom magnification and zoom in to detect characteristic parts of vehicle VCL1 (e.g., vehicle type, color, make, license plate) with high accuracy using AI.

[0104] Furthermore, if the AI ​​detects the vehicle type or color and license plate number of vehicle VCL1 in the image captured by the imaging unit after the camera parameters have been set, the processor 14 performs character recognition of the license plate and matches it against a license plate blacklist. As a result, the surveillance camera 1 can generate a license plate cutout image for matching the owner of vehicle VCL1 by detecting the characteristic parts of vehicle VCL1 in the order of vehicle type, color, and license plate, and based on the matching result, it can notify the monitor in detail of the vehicle VCL1 that has entered the imaging area.

[0105] Although various embodiments have been described above with reference to the drawings, it goes without saying that this disclosure is not limited to such examples. It is clear to those skilled in the art that various modifications, alterations, substitutions, additions, deletions, and equivalents can be conceived within the scope of the claims, and these are also understood to fall within the technical scope of this disclosure. Furthermore, the components of the various embodiments described above can be combined arbitrarily without departing from the spirit of the invention. [Industrial applicability]

[0106] This disclosure provides a useful surveillance camera, parameter determination method, and program that adaptively determine camera parameters that contribute to improving the accuracy of monitoring people's faces or vehicle feature parts in images of the imaging area, thereby accurately improving image quality and streamlining surveillance operations. [Explanation of Symbols]

[0107] 1 Surveillance camera 11 lenses 12 Image sensor 13.53 memory 14, 52 processors 15 AI Processing Unit 15a AI Processing Unit 15b Learning Model Memory 16, 51 Communication IF Circuit 18 External storage medium IF 19 IR Lighting Section 50 matching servers 54 Databases 100 surveillance camera systems LM1 Lens Drive Unit M1 external storage medium PTM1 Pan-Tilt Drive Unit

Claims

1. A surveillance camera equipped with artificial intelligence, An imaging unit that images the imaging area, A detection unit that uses the aforementioned artificial intelligence to detect a person appearing in the captured image of the imaging area, The system includes a control unit that estimates the position of the person's face in the captured image and selectively determines the camera parameters of the surveillance camera based on either the brightness of the face or the movement speed, or both. The control unit, after detecting the person by the process of detecting the presence or absence of the person using the artificial intelligence, but before performing the process of detecting the presence or absence of the person's face, sets the determined camera parameters to the imaging unit. Surveillance camera.

2. The control unit sets the camera parameters. The surveillance camera according to claim 1.

3. If the control unit cannot detect the person in the captured image, it determines and sets the camera parameters so that the person in the captured image captured by the imaging unit can be detected by the detection unit. The surveillance camera according to claim 1.

4. The camera parameter is the exposure time of the imaging unit. The control unit determines the exposure time such that the face of a person in the captured image captured by the imaging unit can be detected by the detection unit. The surveillance camera according to claim 1.

5. The camera parameter is the gain of the imaging unit, The control unit determines the gain such that the face of a person in the image captured by the imaging unit can be detected by the detection unit. The surveillance camera according to claim 1.

6. After detecting the person's face, the control unit determines a tone curve so that the image captured by the imaging unit is suitable for the face matching process. The surveillance camera according to claim 1.

7. The system further includes a communication unit that, after the camera parameters have been set, detects the face of the person in the image captured by the imaging unit and the face detection score is above a threshold, sends a face-extracted image, obtained by extracting the face from the image, to a face matching server that is communicably connected to the surveillance camera. The surveillance camera according to claim 1.

8. A method for determining camera parameters performed by a surveillance camera equipped with artificial intelligence, The steps include capturing an image of the imaging area, The steps include: using the artificial intelligence described above to detect a person appearing in the captured image of the imaging area; The step of estimating the position of the person's face in the captured image and selectively determining the camera parameters of the surveillance camera based on either or both the brightness and / or movement speed of the face, If the presence or absence of the person is detected by the process of detecting the presence or absence of the person using the artificial intelligence, the process further includes the step of setting the determined camera parameters to the surveillance camera after the detection but before the process of detecting the presence or absence of the person's face. Method for determining camera parameters.

9. A computer, a surveillance camera equipped with artificial intelligence, The steps include capturing an image of the imaging area, The steps include: using the artificial intelligence described above to detect a person appearing in the captured image of the imaging area; The process involves estimating the position of the person's face in the captured image and selectively determining the camera parameters of the surveillance camera based on either the brightness or the movement speed of the face, or both. If the presence or absence of the person is detected by the process of detecting the presence or absence of the person using the aforementioned artificial intelligence, the following step is to be performed after the detection but before the process of detecting the presence or absence of the person's face is performed: setting the determined camera parameters to the surveillance camera. program.

10. A surveillance camera equipped with artificial intelligence, An imaging unit that images the imaging area, A detection unit that uses the artificial intelligence to detect vehicles appearing in the captured image of the imaging area, The system includes a control unit that selectively determines camera parameters of the surveillance camera based on either or both the brightness and / or speed of the vehicle in the captured image, The control unit, after detecting the vehicle by the process of detecting the presence or absence of the vehicle using the artificial intelligence, but before performing the process of detecting the vehicle's type or color, sets the determined camera parameters to the imaging unit. Surveillance camera.

11. The control unit sets the camera parameters. The surveillance camera according to claim 10.

12. If the vehicle cannot be detected in the captured image, the control unit determines and sets the camera parameters so that the vehicle captured in the image taken by the imaging unit can be detected by the detection unit. The surveillance camera according to claim 10.

13. The camera parameter is the exposure time of the imaging unit. The control unit determines the exposure time such that the vehicle type, color, vehicle model, or license plate of a vehicle captured in the image captured by the imaging unit can be detected by the detection unit. The surveillance camera according to claim 10.

14. The camera parameter is the gain of the imaging unit, The control unit determines the gain such that the vehicle type, vehicle color, vehicle model, or license plate visible in the image captured by the imaging unit can be detected by the detection unit. The surveillance camera according to claim 10.

15. After detecting the vehicle type or color, the control unit determines at least one of the pan rotation and tilt rotation amounts of the surveillance camera so that the image captured by the imaging unit is suitable for recognizing the license plate number included in the image. The surveillance camera according to claim 10.

16. The control unit, after detecting the vehicle type or color, determines the zoom magnification of the imaging unit so that the image captured by the imaging unit is suitable for recognizing the license plate number included in the image. The surveillance camera according to claim 10.

17. The system further includes a character recognition unit that, after the camera parameters have been set, performs character recognition of the license plate and compares it with a blacklist of license plates when the detection unit detects the vehicle type or color and the license plate number in the captured image taken by the imaging unit. A surveillance camera according to any one of claims 13 to 16.

18. A method for determining camera parameters performed by a surveillance camera equipped with artificial intelligence, The steps include capturing an image of the imaging area, The steps include using the artificial intelligence to detect a vehicle in the image captured in the image of the imaging area, The step of selectively determining camera parameters of the surveillance camera based on either or both the brightness and / or speed of the vehicle in the captured image, If the presence or absence of the vehicle is detected by the process of detecting the presence or absence of the vehicle using the artificial intelligence, the process further includes the step of setting the determined camera parameters to the surveillance camera after the detection but before the process of detecting the vehicle's type or color. Method for determining camera parameters.

19. A computer, a surveillance camera equipped with artificial intelligence, The steps include capturing an image of the imaging area, The steps include using the artificial intelligence to detect a vehicle in the image captured in the image of the imaging area, The steps include: selectively determining the camera parameters of the surveillance camera based on either or both the brightness and / or speed of the vehicle in the captured image; If the presence or absence of the vehicle is detected by the process of detecting the presence or absence of the vehicle using the aforementioned artificial intelligence, the following step is to be performed after the detection but before the process of detecting the vehicle type or color of the vehicle is performed: setting the determined camera parameters to the surveillance camera. program.

Images