Visual monitoring method and system

Abstract

The invention discloses a visual monitoring method and system. The method comprises the following steps: calling a first camera to shoot a panoramic image; performing target detection on the panoramic image based on an AI algorithm to obtain target information of each target contained in the panoramic image; determining the target as a target to be subjected to secondary identification; for each target to be subjected to secondary identification, determining a rotation angle of a camera holder where the second camera is located according to the coordinate information and the depth information corresponding to the target to be subjected to secondary identification; pulse parameters of the servo motor are generated according to the rotation angle; adjusting the shooting angle of the second camera according to a pulse parameter for shooting to obtain a target amplified image; and target detection is performed on the target amplified image based on the AI algorithm to obtain a target detection result, so that the accuracy and comprehensiveness of target recognition can be improved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a visual monitoring method and system. Background Technology

[0002] Camera surveillance systems are an important component of modern security management. These systems use AI algorithms to analyze video content in real time, extracting useful information such as faces and license plate numbers for monitoring, alarm functions, and historical video playback. They can be widely used in public safety, traffic monitoring, and commercial settings, enabling real-time detection and identification of potential security risks.

[0003] Existing camera surveillance systems suffer from low camera resolution, complex shooting scenes, and some targets being too small in the images, leading to numerous false positives and false negatives during subsequent AI image recognition. Therefore, there is an urgent need for those skilled in the art to provide a camera surveillance system capable of accurately and comprehensively identifying targets in captured images. Summary of the Invention

[0004] The purpose of this invention is to provide a visual monitoring method and system that can solve the problem in the prior art of being unable to accurately and comprehensively identify targets in captured images.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] This invention provides a visual monitoring method, wherein the method is applied to a camera monitoring system including a first camera and a second camera, and the method includes:

[0007] Use the first camera to capture a panoramic image;

[0008] The panoramic image is subjected to target detection based on AI algorithm to obtain target information of each target contained in the panoramic image, wherein the target information includes coordinate information and depth information;

[0009] The target is identified as the target to be identified a second time.

[0010] For each target to be identified a second time, the rotation angle of the camera gimbal where the second camera is located is determined based on the coordinate information and depth information corresponding to the target to be identified a second time.

[0011] The pulse parameters of the servo motor are generated based on the rotation angle;

[0012] The shooting angle of the second camera is adjusted according to the pulse parameters to capture a magnified image of the target.

[0013] perform target detection on the target zoomed-in image based on the AI algorithm to obtain a target detection result.

[0014] Optionally, for each of the to-be-secondly-identified targets, the step of determining the rotation angle of the camera holder where the second camera is located according to the coordinate information and the depth information corresponding to the to-be-secondly-identified target comprises:

[0015] for each of the to-be-secondly-identified targets, the depth information corresponding to the to-be-secondly-identified target and the first camera intrinsic matrix are used to calculate the three-dimensional coordinates of the to-be-secondly-identified target under the first camera;

[0016] a rotation matrix and a translation vector are used to generate a conversion matrix;

[0017] the conversion matrix is used to convert the three-dimensional coordinates under the first camera into three-dimensional coordinates under the second camera;

[0018] the three-dimensional coordinates under the second camera are used to calculate the rotation angle of the camera holder where the second camera is located.

[0019] Optionally, the step of calculating the rotation angle of the camera holder where the second camera is located according to the three-dimensional coordinates under the second camera comprises:

[0020] the three-dimensional coordinates under the second camera and the current holder pitch angle are used to calculate the target holder pitch angle;

[0021] the three-dimensional coordinates under the second camera and the current holder yaw angle are used to calculate the target holder yaw angle.

[0022] Optionally, after the step of performing target detection on the panoramic image based on the AI algorithm to obtain the target information of each target contained in the panoramic image, the method further comprises:

[0023] a missed target that is not detected in the panoramic image is determined;

[0024] target information corresponding to the missed target is determined;

[0025] the missed target is determined as a to-be-secondly-identified target.

[0026] Optionally, the first camera is an RGB-D depth camera, and the second camera is a long-focus camera.

[0027] Optionally, the step of performing target detection on the panoramic image based on the AI algorithm to obtain the target information of each target contained in the panoramic image comprises:

[0028] The AI algorithm is used for target recognition on the panoramic image to determine a target frame contained in the panoramic image;

[0029] For each target frame, the center point coordinate and the boundary frame information of the target frame on the image are obtained to determine a target region;

[0030] The depth information of the target region is extracted from the depth map corresponding to the panoramic image.

[0031] The embodiment of the application further provides a visual monitoring system, wherein the system comprises a first camera, a second camera and a camera holder.

[0032] A calling module is configured to call the first camera to shoot a panoramic image.

[0033] A first detection module is configured to perform target detection on the panoramic image based on an AI algorithm to obtain target information of each target contained in the panoramic image, wherein the target information comprises coordinate information and depth information.

[0034] A first determination module is configured to determine the target as a target to be identified again.

[0035] An angle determination module is configured to determine, for each target to be identified again, a rotation angle of the camera holder of the second camera according to the coordinate information and the depth information corresponding to the target.

[0036] A parameter generation module is configured to generate a pulse parameter of a servo motor according to the rotation angle.

[0037] An adjustment module is configured to adjust the shooting angle of the second camera to shoot a target magnified image according to the pulse parameter.

[0038] A second detection module is configured to perform target detection on the target magnified image based on the AI algorithm to obtain a target detection result.

[0039] Optionally, the angle determination module comprises:

[0040] A first sub-module is configured to calculate, for each target to be identified again, a three-dimensional coordinate of the target to be identified again under the first camera according to the depth information corresponding to the target to be identified again and the intrinsic matrix of the first camera.

[0041] A second sub-module is configured to generate a conversion matrix according to a rotation matrix and a translation vector.

[0042] A third sub-module is configured to convert the three-dimensional coordinate under the first camera into a three-dimensional coordinate under the second camera by using the conversion matrix.

[0043] a fourth sub-module configured to calculate a rotation angle of a camera holder on which the second camera is located according to the three-dimensional coordinates under the second camera.

[0044] Optionally, the fourth sub-module is specifically configured to:

[0045] calculate a target holder pitch angle according to the three-dimensional coordinates under the second camera and a current holder pitch angle;

[0046] calculate a target holder yaw angle according to the three-dimensional coordinates under the second camera and a current holder yaw angle.

[0047] Optionally, the camera holder further comprises:

[0048] a missed target determination module configured to, after the first detection module performs target detection on the panoramic image based on an AI algorithm to obtain target information of each target contained in the panoramic image, determine a missed target in the panoramic image that is not detected;

[0049] an information determination module configured to determine target information corresponding to the missed target;

[0050] a second determination module configured to determine the missed target as a target to be identified again.

[0051] Optionally, the first camera is an RGB-D depth camera, and the second camera is a long-focus camera.

[0052] Optionally, the first detection module comprises:

[0053] a fifth sub-module configured to perform target identification on the panoramic image based on an AI algorithm to determine a target frame contained in the panoramic image;

[0054] a sixth sub-module configured to, for each target frame, obtain a center point coordinate and bounding box information of the target frame on an image to determine a target region;

[0055] a seventh sub-module configured to extract depth information of the target region from a depth map corresponding to the panoramic image.

[0056] The visual monitoring solution provided in this application sets up a first camera and a second camera in a camera monitoring system. During visual monitoring, the first camera captures a panoramic image; an AI algorithm is used to perform target detection on the panoramic image to obtain target information for each target contained in the panoramic image; the target is identified as a target to be identified a second time; for each target to be identified a second time, the rotation angle of the camera pan-tilt unit of the second camera is determined based on the coordinate and depth information corresponding to the target; pulse parameters of the servo motor are generated based on the rotation angle; the shooting angle of the second camera is adjusted based on the pulse parameters to capture a magnified image of the target; and the AI ​​algorithm is used to perform target detection on the magnified image to obtain the target detection result. This solution, on the one hand, improves the accuracy and comprehensiveness of target identification by using the second camera to capture a magnified image of the target after detecting a target in the panoramic image captured by the first camera, thus enhancing the accuracy and comprehensiveness of target identification; on the other hand, the second camera's capture does not affect the continuous monitoring of the first camera, ensuring the timeliness and continuity of monitoring. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating the steps of a visual monitoring method according to an embodiment of this application;

[0058] Figure 2 This is a schematic diagram illustrating the structure of a camera monitoring system according to an embodiment of this application;

[0059] Figure 3 This is a structural block diagram illustrating a functional module in a camera gimbal according to an embodiment of this application. Detailed Implementation

[0060] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0061] The visual monitoring solution provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0062] As attached Figure 1 As shown, the visual monitoring method of this application embodiment includes the following steps:

[0063] Step 101: Use the first camera to capture a panoramic image.

[0064] The visual monitoring method provided in the embodiments of the present application can be applied to a camera monitoring system comprising a first camera and a second camera. The camera monitoring system further comprises a camera holder, and a processor is arranged in the camera holder to control the first camera and the second camera to execute the visual monitoring method shown in the embodiments of the present application. The camera holder comprises the processor and a storage medium. The storage medium stores a visual monitoring related computer program, and the processor in the camera holder executes the computer program to implement the visual monitoring process.

[0065] In an optional embodiment, the first camera is an RGB-D depth camera, and the second camera is a long-focus camera.

[0066] An exemplary structural diagram of a camera monitoring system is shown in FIG. 1. Figure 2 As shown in FIG. 1, the camera monitoring system comprises an RGB-D depth camera 1, a camera holder 2, a holder system 3 composed of two axes, a long-focus camera 4, and a long-focus camera shell 5. Figure 2 The holder system 3 in FIG. 1 can be regarded as a camera holder.

[0067] It should be noted that, Figure 2 The structural diagram of the camera monitoring system shown in FIG. 1 is exemplary only, and in actual implementation, the specific model of the first camera and the second camera can be set by those skilled in the art according to actual needs, and the embodiments of the present application do not make specific limitations thereon.

[0068] The embodiments of the present application are described on the basis that the precondition of camera system parameter calibration has been completed. In actual implementation, camera system parameter calibration needs to be performed before step 101. Camera system parameter calibration can be performed in the following manner:

[0069] RGB-D camera intrinsic parameter calibration: obtain the parameters and distortion coefficients of the depth map and the RGB image;

[0070] Master (RGB-D camera)-slave (long-focus camera) extrinsic parameter calibration: a checkerboard combined with a Charuco board can be used to improve the accuracy and calculate the rotation matrix R and the translation vector t. The Charuco board is a camera calibration board.

[0071] After the camera system parameter calibration is completed, the camera is erected to perform normal video shooting, and the first camera can be used to shoot panoramic images.

[0072] Step 102: performing target detection on the panoramic image based on an AI algorithm to obtain target information of each target contained in the panoramic image.

[0073] The target information comprises coordinate information and depth information.

[0074] In actual implementation, an AI algorithm detection interval time can be set, one frame of panoramic image in the interval time is input to the AI model for inference, and the target contained in the panoramic image is recognized. The AI model is pre-set with an AI algorithm, and target detection is performed by executing the AI algorithm. It should be noted that the AI algorithm can adopt any appropriate algorithm, and the specific AI algorithm adopted in the embodiments of the present application is not limited, and it can only be ensured that it has the ability to recognize targets in images. The interval time can be flexibly set by those skilled in the art, and the embodiments of the present application do not make specific limitations thereon, for example, it can be set to 1 second, 2 seconds or 3 seconds, etc.

[0075] In an alternative embodiment, the way of detecting targets in the panoramic image based on the AI algorithm to obtain the target information of each target contained in the panoramic image can include the following sub-steps:

[0076] Sub-step 1: based on the AI algorithm, the target in the panoramic image is recognized to determine the target frame contained in the panoramic image;

[0077] Sub-step 2: for each target frame, the center point coordinates and the bounding box information of the target frame on the image are obtained to determine the target region;

[0078] Sub-step 3: the depth information of the target region is extracted from the depth map corresponding to the panoramic image.

[0079] Each panoramic image corresponds to a depth map, so when a panoramic image is detected, its corresponding depth map can also be obtained.

[0080] Step 103: determine the detected target as a target to be identified again.

[0081] In an alternative embodiment, the way of detecting targets in the panoramic image based on the AI algorithm to obtain the target information of each target contained in the panoramic image can include the following flow:

[0082] Sub-step 1: determine the undetected missed target in the panoramic image;

[0083] Sub-step 2: determine the target information corresponding to the missed target;

[0084] Sub-step 3: determine the missed target as a target to be identified again.

[0085] Determining the missed target as a target to be identified again can avoid target missing.

[0086] The first camera performs panoramic shooting on a monitoring scene, target detection is performed on the shot picture using an AI detection technology, coordinate conversion is performed on the detected target or a target not detected at a certain position, and the second camera is used to shoot the target position to enlarge the target and further confirm the real situation of the target. Therefore, the target to be identified again needs to be determined in this step.

[0087] For the case where multiple targets to be identified again exist at the same time, different measurement strategies can be set to set priorities to achieve different levels of target monitoring. For example, the target that should appear but is not detected by the first AI detection is detected first, and then the target detected by the first AI detection is detected. For another example, the second detection priority is set according to the target type. The priority strategy is not specifically limited in the embodiments of the application.

[0088] Step 104: For each target to be identified again, the rotation angle of the camera holder where the second camera is located is determined according to the coordinate information and depth information corresponding to the target to be identified again.

[0089] In an optional embodiment, for each target to be identified again, the rotation angle of the camera holder where the second camera is located can be determined according to the coordinate information and depth information corresponding to the target to be identified again in the following sub-steps:

[0090] Sub-step 1: For each target to be identified again, the three-dimensional coordinates of the target to be identified again under the first camera are calculated according to the depth information corresponding to the target to be identified again and the intrinsic matrix of the first camera.

[0091] The 3D coordinates of the target under the main camera (i.e., the first camera) are calculated as follows:

[0092] Specifically, the inverse projection calculation can be performed through the camera model:

[0093]

[0094] wherein D depth represents the depth value of the target point, and the unit is meter; K depth represents the intrinsic matrix of the depth lens (i.e., the first camera).

[0095] Sub-step 2: A conversion matrix is generated according to the rotation matrix and the translation vector.

[0096] Sub-step 3: The three-dimensional coordinates under the first camera are converted into three-dimensional coordinates under the second camera using the conversion matrix.

[0097] The 3D coordinates under the main camera are converted into the coordinate system under the slave camera (i.e., the second camera) as follows:

[0098]

[0099] wherein: R represents a rotation matrix, and t represents a translation vector.

[0100] Sub-step 4: According to the three-dimensional coordinates under the second camera, the rotation angle of the camera holder where the second camera is located is calculated.

[0101] A feasible implementation manner of calculating the rotation angle of the camera holder where the second camera is located according to the three-dimensional coordinates under the second camera can be as follows:

[0102] According to the three-dimensional coordinates under the second camera and the current tilt angle of the holder, the target tilt angle of the holder is calculated; and according to the three-dimensional coordinates under the second camera and the current yaw angle of the holder, the target yaw angle of the holder is calculated.

[0103] Specifically, the target tilt angle of the holder can be calculated based on the following formula:

[0104]

[0105] The target yaw angle of the holder can be calculated based on the following formula:

[0106]

[0107] wherein θ1 represents the target tilt angle of the holder, and θ0 represents the current tilt angle of the holder. represents the target yaw angle of the holder, represents the current yaw angle of the holder.

[0108] Step 105: Pulse parameters of a servo motor are generated according to the rotation angle.

[0109] The angle is converted into control pulses of the holder servo motor, so that the camera holder is rotated to adjust the shooting angle of the second camera.

[0110] Step 106: The shooting angle of the second camera is adjusted according to the pulse parameters to shoot a target magnified image.

[0111] The target magnified image is shot by the second camera, and the target information shot by the magnified image is more comprehensive and accurate. Based on the target magnified image, AI detection is performed again, so that more accurate and comprehensive target information can be obtained.

[0112] Step 107: Target detection is performed on the target magnified image based on an AI algorithm to obtain a target detection result.

[0113] It should be noted that steps 104-107 are the process of secondary AI detection of one target to be identified again. In actual implementation process, if multiple targets to be identified again are included, the detection order of the targets to be identified again can be sorted according to the pre-set priority strategy, and steps 104-107 are repeatedly executed to detect each target to be identified again.

[0114] The visual monitoring method provided in the embodiments of the application sets a first camera and a second camera in a camera monitoring system, calls the first camera to shoot a panoramic image when visual monitoring is performed, performs target detection on the panoramic image based on an AI algorithm to obtain target information of each target included in the panoramic image, determines the target as a target to be identified again, determines a rotation angle of a camera holder where the second camera is located according to coordinate information and depth information corresponding to each target to be identified again for each target to be identified again, generates pulse parameters of a servo motor according to the rotation angle, adjusts a shooting angle of the second camera to shoot according to the pulse parameters to obtain an enlarged target image, and performs target detection on the enlarged target image based on the AI algorithm to obtain a target detection result. The method provided in the application can improve the accuracy and comprehensiveness of target identification by calling the second camera to shoot the target position and enlarge the shooting target when the target is detected by the first camera to shoot the panoramic image, and further confirm the real situation of the target, and can ensure the timeliness and continuity of monitoring because the second camera does not affect the continuous monitoring of the first camera when shooting.

[0115] Figure 3 To realize the structure block diagram of the functional modules in the camera holder in the visual monitoring system according to an embodiment of the application.

[0116] The visual monitoring system provided in the embodiments of the application includes a first camera, a second camera, and a camera holder, and the camera holder includes the following functional modules.

[0117] The calling module 301 is configured to call the first camera to shoot a panoramic image.

[0118] The first detection module 302 is configured to perform target detection on the panoramic image based on an AI algorithm to obtain target information of each target included in the panoramic image, wherein the target information includes coordinate information and depth information.

[0119] The first determination module 303 is configured to determine the target as a target to be identified again.

[0120] The angle determination module 304 is configured to determine a rotation angle of a camera holder where the second camera is located according to coordinate information and depth information corresponding to each target to be identified again for each target to be identified again.

[0121] The parameter generation module 305 is configured to generate a pulse parameter of the servo motor according to the rotation angle.

[0122] The adjustment module 306 is configured to adjust a shooting angle of the second camera to shoot according to the pulse parameter, to obtain a target zoomed image.

[0123] The second detection module 307 is configured to perform target detection on the target zoomed image based on the AI algorithm, to obtain a target detection result.

[0124] Optionally, the angle determination module comprises:

[0125] The first sub-module is configured to, for each of the to-be-recognized target, calculate a three-dimensional coordinate of the to-be-recognized target under the first camera according to the depth information corresponding to the to-be-recognized target and the first camera intrinsic parameter matrix.

[0126] The second sub-module is configured to generate a conversion matrix according to the rotation matrix and the translation vector.

[0127] The third sub-module is configured to convert the three-dimensional coordinate under the first camera into a three-dimensional coordinate under the second camera by using the conversion matrix.

[0128] The fourth sub-module is configured to calculate a rotation angle of a camera holder of the second camera according to the three-dimensional coordinate under the second camera.

[0129] Optionally, the fourth sub-module is specifically configured to:

[0130] calculate a target holder pitch angle according to the three-dimensional coordinate under the second camera and a current holder pitch angle;

[0131] calculate a target holder yaw angle according to the three-dimensional coordinate under the second camera and a current holder yaw angle.

[0132] Optionally, the camera holder further comprises:

[0133] The missed target determination module is configured to, after the first detection module performs target detection on the panoramic image based on the AI algorithm to obtain target information of each target contained in the panoramic image, determine a missed target that is not detected in the panoramic image.

[0134] The information determination module is configured to determine target information corresponding to the missed target.

[0135] The second determination module is configured to determine the missed target as a to-be-recognized target.

[0136] Optionally, the first camera is an RGB-D depth camera, and the second camera is a long-focus camera.

[0137] Optionally, the first detection module comprises:

[0138] a fifth sub-module configured to perform target recognition on the panoramic image based on an AI algorithm to determine a target frame contained in the panoramic image;

[0139] a sixth sub-module configured to, for each target frame, acquire a center point coordinate and a bounding box information of the target frame on the image to determine a target region;

[0140] a seventh sub-module configured to extract depth information of the target region from a depth map corresponding to the panoramic image.

[0141] The visual monitoring system provided by the embodiments of the present application can achieve Figure 3 The method embodiments achieve various processes, and thus repeated descriptions are omitted. Figure 1

[0142] The visual monitoring system provided by the embodiments of the present application comprises a first camera, a second camera, and a camera holder. When performing visual monitoring, the camera holder calls the first camera to capture a panoramic image. Target detection is performed on the panoramic image based on an AI algorithm to obtain target information of each target contained in the panoramic image. The target is determined as a target to be recognized again. For each target to be recognized again, a rotation angle of the camera holder where the second camera is located is determined according to coordinate information and depth information corresponding to the target to be recognized again. Pulse parameters of a servo motor are generated according to the rotation angle. The second camera is adjusted to capture an image according to the pulse parameters, and a target magnified image is obtained. Target detection is performed on the target magnified image based on the AI algorithm to obtain a target detection result. On the one hand, when a target is detected by the panoramic image captured by the first camera, the second camera is called to capture the target position, the target is magnified, and the real situation of the target is further confirmed, which can improve the accuracy and comprehensiveness of target recognition. On the other hand, the second camera does not affect the continuous monitoring of the first camera, and the timeliness and continuity of monitoring can be ensured.

[0143] Optionally, the embodiments of the present application further provide an electronic device comprising a processor, a memory, a program or instructions stored on the memory and executable on the processor. The program or instructions are executed by the processor to implement various processes performed by the camera holder and achieve the same technical effects. Thus, repeated descriptions are omitted.

[0144] The memory is configured to store a computer program.

[0145] ​The processor is configured to implement the visual monitoring method performed by the processor in the above method embodiments when executing the program stored in the memory.

[0146] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or only one type of bus.

[0147] The communication interface is used for communication between the terminal and other devices.

[0148] The memory can include a Random Access Memory (RAM) and can also include a non-volatile memory, such as at least one disk memory. Optionally, the memory can also be at least one storage device located away from the aforementioned processor.

[0149] The processor mentioned above can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; can also be a Digital Signal Processing (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0150] It should be noted that in this document, the terms "comprise", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or device. Without more limitations, the element defined by the statement "comprises a" does not exclude the presence of another identical element in the process, method, article or device that includes the element.

[0151] The above is the preferred embodiment of the present application, it should be noted that for those skilled in the art, without departing from the principles described in the present application, can also make several improvements and refinements, these improvements and refinements should also be considered as the scope of protection of the present application.

Claims

1. A method of visual monitoring, characterized by, The method is applied to a camera monitoring system comprising a first camera and a second camera, and the method comprises: calling the first camera to take a panoramic image; performing target detection on the panoramic image based on an AI algorithm to obtain target information of each target contained in the panoramic image, wherein the target information comprises coordinate information and depth information; determining the target as a target to be identified again; for each target to be identified again, determining a rotation angle of a camera holder where the second camera is located according to the coordinate information and the depth information corresponding to the target to be identified again; generating pulse parameters of a servo motor according to the rotation angle; adjusting the shooting angle of the second camera to take a target zoomed-in image according to the pulse parameters; performing target detection on the target zoomed-in image based on the AI algorithm to obtain a target detection result.

2. The method of claim 1, wherein, For each target to be identified again, the step of determining a rotation angle of a camera holder where the second camera is located according to the coordinate information and the depth information corresponding to the target to be identified again comprises: for each target to be identified again, calculating three-dimensional coordinates of the target to be identified again under the first camera according to the depth information corresponding to the target to be identified again and an intrinsic matrix of the first camera; generating a conversion matrix according to a rotation matrix and a translation vector; converting the three-dimensional coordinates under the first camera into three-dimensional coordinates under the second camera by using the conversion matrix; calculating the rotation angle of the camera holder where the second camera is located according to the three-dimensional coordinates under the second camera.

3. The method of claim 2, wherein, The step of calculating the rotation angle of the camera holder where the second camera is located according to the three-dimensional coordinates under the second camera comprises: calculating a target holder pitch angle according to the three-dimensional coordinates under the second camera and a current holder pitch angle; calculating a target holder yaw angle according to the three-dimensional coordinates under the second camera and a current holder yaw angle.

4. The method of claim 1, wherein, After the step of performing target detection on the panoramic image based on the AI algorithm to obtain target information of each target contained in the panoramic image, the method further comprises: determining a missed target that is not detected in the panoramic image; determining target information corresponding to the missed target; determining the missed target as a target to be identified again.

5. The method of claim 1, wherein, The first camera is an RGB-D depth camera, and the second camera is a long-focus camera.

6. The method of claim 1, wherein, The step of performing target detection on the panoramic image based on the AI algorithm to obtain target information of each target contained in the panoramic image comprises: performing target identification on the panoramic image based on the AI algorithm to determine a target frame contained in the panoramic image; for each target frame, obtaining a center point coordinate and boundary box information of the target frame on an image to determine a target region; extracting depth information of the target region from a depth map corresponding to the panoramic image.

7. A vision monitoring system characterized by, The system comprises a first camera, a second camera, and a camera holder. The camera holder comprises: a calling module configured to call the first camera to take a panoramic image; The first detection module is configured to perform target detection on the panoramic image based on an AI algorithm to obtain target information of each target contained in the panoramic image, wherein the target information comprises coordinate information and depth information. The first determination module is configured to determine the target as a target to be identified again. The angle determination module is configured to determine, for each target to be identified again, a rotation angle of a camera holder where the second camera is located according to the coordinate information and the depth information corresponding to the target to be identified again. The parameter generation module is configured to generate a pulse parameter of a servo motor according to the rotation angle. The adjustment module is configured to adjust a shooting angle of the second camera to perform shooting according to the pulse parameter, and obtain a target zoomed image. The second detection module is configured to perform target detection on the target zoomed image based on the AI algorithm to obtain a target detection result.

8. The system of claim 7, wherein, The angle determination module comprises: The first sub-module is configured to calculate, for each target to be identified again, a three-dimensional coordinate of the target to be identified again under the first camera according to the depth information corresponding to the target to be identified again and an intrinsic matrix of the first camera. The second sub-module is configured to generate a conversion matrix according to a rotation matrix and a translation vector. The third sub-module is configured to convert the three-dimensional coordinate under the first camera into a three-dimensional coordinate under the second camera by using the conversion matrix. The fourth sub-module is configured to calculate a rotation angle of a camera holder where the second camera is located according to the three-dimensional coordinate under the second camera.

9. The system of claim 8, wherein, The fourth sub-module is specifically configured to: calculate a target camera tilt angle according to the three-dimensional coordinate under the second camera and a current camera tilt angle; and calculate a target camera yaw angle according to the three-dimensional coordinate under the second camera and a current camera yaw angle.

10. The system of claim 7, wherein, The camera holder further comprises: The undetected target determination module is configured to determine an undetected target in the panoramic image after the first detection module performs target detection on the panoramic image based on the AI algorithm to obtain target information of each target contained in the panoramic image. The information determination module is configured to determine target information corresponding to the undetected target. The second determination module is configured to determine the undetected target as a target to be identified again.

11. The system of claim 7, wherein, The first camera is an RGB-D depth camera, and the second camera is a long-focus camera.

12. The system of claim 7, wherein, The first detection module comprises: The fifth sub-module is configured to perform target identification on the panoramic image based on the AI algorithm to determine a target frame contained in the panoramic image. The sixth sub-module is configured to obtain a center point coordinate and bounding box information of each target frame on an image to determine a target region. The seventh sub-module is configured to extract depth information of the target region from a depth map corresponding to the panoramic image.

Images