Personnel monitoring method and system for power distribution room based on multi-camera joint tracking algorithm
Through multi-camera joint tracking algorithm and artificial intelligence technology, real-time and accurate tracking and behavioral monitoring of distribution room personnel are achieved, solving the problems of blind spots and high cost in the existing technology, and improving monitoring accuracy and security.
Patent Information
- Application Number
- CN202410429837.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-10
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-04-10
AI Technical Summary
The existing distribution room personnel monitoring technology is difficult to achieve real-time and accurate tracking and behavioral monitoring of individual patrol personnel, and the cost is high, so it cannot effectively reduce the number of cameras to reduce tracking blind spots.
The camera is initialized through artificial intelligence recognition technology, and the target personnel are tracked adaptively, the camera is used to calculate the precise position of the target personnel using camera offset and world coordinates, and the camera magnification is automatically adjusted to identify behavior and generate early warning information.
Real-time, blind spot tracking and monitoring of personnel in the distribution room is realized, the number of cameras is reduced, the tracking accuracy and behavior recognition accuracy are improved, the cost is reduced, and the personal safety of inspectors is ensured.
Smart Images

Figure CN118317188B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of intelligent tracking of personnel in a distribution room, and specifically to a method and system for monitoring personnel in a distribution room based on a multi-camera joint tracking algorithm. Background Art
[0002] At present, the safety management of power distribution rooms and switch rooms is mainly achieved by manual monitoring. However, due to the large number, large scope and strong closedness of power distribution rooms and switch rooms, it is difficult for manual monitoring and tracking of the status of individual inspectors in the power distribution room in real time, so as to timely discover individual inspectors with abnormal behavior or illegal operations and ensure the personal safety of inspectors.
[0003] In order to solve this problem, most existing technologies deploy multiple cameras so that the camera images cover all areas, and use some personnel tracking algorithms to track, thereby achieving tracking and monitoring of patrol personnel. However, this traditional personnel tracking method cannot control the movement of the camera, and the tracking method of multiple cameras with fixed fields of view increases the tracking cost, and it is impossible to obtain the precise location of the personnel and the personnel's patrol route. Summary of the invention
[0004] In order to achieve real-time tracking and monitoring of distribution room personnel, improve tracking accuracy and reduce tracking costs, the present application provides a distribution room personnel monitoring method based on a multi-camera joint tracking algorithm.
[0005] In a first aspect, the present application provides a method for monitoring personnel in a power distribution room based on a multi-camera joint tracking algorithm, comprising:
[0006] Initialize all surveillance cameras in the power distribution room to the watch position, where the watch position is the set horizontal angle and vertical angle;
[0007] Artificial intelligence recognition technology is used to identify people in the image data collected by all cameras. If a single person is identified, the single-person tracking mode is triggered and the single person is determined as the target person;
[0008] When the single-person tracking mode is triggered, the camera closest to the target person is controlled to adaptively track so that the target person is always in the center of the captured image until the target person leaves the camera's captured image; during the camera adaptive tracking process, the offset of the camera that recognizes the target person is calculated in real time according to the target person's pixels, and the horizontal and vertical viewing angles after the camera is offset are obtained; based on the horizontal and vertical viewing angles after the camera is offset and the world coordinates of the camera, the world coordinates corresponding to the target person's pixels, i.e., the world coordinates of the target person, are calculated in real time;
[0009] When the target person leaves the currently controlled camera shooting screen, the closest camera that recognizes the same target person is switched to continue adaptive tracking; the camera switching and adaptive tracking of the target person are repeated until all cameras fail to recognize the same target person and stop tracking the target person.
[0010] By adopting the above solution, the camera that has identified a single target person is controlled to move continuously to achieve blind spot shooting and personnel tracking in the distribution room, reduce the number of cameras and reduce the tracking cost; the camera offset is calculated according to the pixel coordinates of the target person, and the world coordinates of the target person are calculated in combination with the camera world coordinates to obtain a more accurate position of the target person.
[0011] Preferably, the specific calculation formula for calculating the offset of the camera that identifies the target person in real time based on the target person pixel and calculating the world coordinates corresponding to the target person pixel in real time based on the camera offset and the camera's world coordinates includes:
[0012]
[0013]
[0014] p1=p+delt_x
[0015] t1=t+delt_y
[0016] Where, delt_x is the offset of the camera horizontal angle, delt_y is the offset of the camera vertical angle; (px, py) is the pixel of the target person; im_w is the width of the image taken by the camera; im_h is the height of the image taken by the camera; fov_w is the horizontal field of view of the camera before the camera offset; fov_h is the vertical field of view of the camera before the camera offset; p1 is the horizontal angle of the camera after the camera offset, p is the horizontal angle of the camera before the camera offset, t1 is the vertical angle of the camera after the camera offset, and t is the vertical angle of the camera before the camera offset;
[0017] c_p_x=word_x+abs(tan(deg2rad(t1))*word_z) Where,
[0018] c_p_y=word_y
[0019] a_x=
[0020] (c_p_x-word_x)×cos(deg2rad(p1))-(c_p_y-word_y)×sin(deg2rad(p1))+word_x
[0021] a_y=
[0022] (c_p_x-word_x)×sin(deg2rad(p1))+(c_p_y-word_y)×cos(deg2rad(p1))+word_y
[0023] (a_x, a_y, a_z) is the world coordinate of the target person's pixel; (word_x, word_y, word_z) is the world coordinate of the camera.
[0024] By adopting the above scheme, the camera offset is calculated according to the camera viewing field and the pixel coordinates of the target person, and the world coordinates of the target person corresponding to the pixel coordinates of the target person are calculated in combination with the camera world coordinates, so as to obtain a more accurate position of the target person.
[0025] Preferably, it also includes:
[0026] After controlling the camera to adaptively track the target person so that the target person is always in the center of the captured image, the camera magnification is adjusted according to the proportion of the target person in the picture so that the pixel height of the target person is kept within a preset percentage value of the total pixel height of the camera shooting picture;
[0027] Artificial intelligence recognition technology is used to identify the target person's behavior in the picture taken after the camera magnification is adjusted. When the preset behavior is identified, an early warning message is generated and played in real time; the preset behaviors include: climbing, falling, and improper dressing.
[0028] By adopting the above solution and adjusting the automatic control of the camera magnification, a clearer image of the person can be obtained, providing a basis for accurate identification of the person's behavior, thereby timely and accurately identifying the dangerous behavior of the target person and giving warnings.
[0029] Preferably, it also includes:
[0030] In the process of camera adaptive tracking of the identified target person, the world coordinates of the current target person are calculated based on the pixel coordinates of the currently identified target person;
[0031] Calculate the horizontal angle and vertical angle of the camera after it moves when the center of the camera screen tracks the current target person, and calculate the camera movement time in combination with the preset rate of camera movement; the specific formulas for the horizontal angle and vertical angle of the camera after it moves are:
[0032] t2=rad2deg(arctan(dis / z2))
[0033]
[0034] In the formula, the calculated world coordinates of the target person are (x1, y1, z1), and the known world coordinates of the camera are (x2, y2, z2);
[0035] Obtain the moving speed and direction of the target person based on historical image information, and calculate the current world coordinates of the target person to estimate the world coordinates of the target person after the camera moves;
[0036] When the center of the camera screen tracks the target person after the camera moves, the horizontal angle and vertical angle of the camera after the camera moves are calculated, and tracking is performed based on the horizontal angle and vertical angle of the camera after the camera moves.
[0037] By adopting the above solution, considering that the camera obtains the target person's pixels to calculate the offset and move with a lag, the target person will move in real time. After the camera moves according to the calculated offset, it may not be able to cover the target person, resulting in tracking failure; the speed of the target person is obtained, the world coordinates of the target person after the camera moves are estimated, and the world coordinates of the target person after the camera moves are used for tracking to ensure the timeliness of tracking.
[0038] Preferably, the controlling distance to identify the nearest camera adaptive tracking of the target person includes:
[0039] If there are multiple cameras that recognize the target person at the same time, determine whether the recognized target person belongs to the same target person; for multiple cameras that recognize the same target person at the same time, calculate the world coordinates of the current target person according to the pixel coordinates of the target person recognized by each camera, calculate the distance between the camera and the target person in combination with the world coordinates of each camera, and select the camera that recognizes the target person closest to it as the control object;
[0040] For multiple cameras that recognize different target persons at the same time, the camera that recognizes the corresponding target person at the closest control distance is used as the control object for each independent target person.
[0041] By adopting the above solution, taking into account the situation where multiple cameras recognize the same target person at the same time and multiple cameras recognize different target persons, a target person is set to correspond to a camera that is closest to the target person and adaptive tracking control is performed until switching to the next camera to ensure orderly tracking.
[0042] Preferably, making the target person always in the center of the captured image means that the waist of the target person is always in the center of the captured image.
[0043] By adopting the above solution, the camera offset is determined by waist pixels, thereby ensuring that the world coordinates of the target person obtained by calculation are more accurate.
[0044] Preferably, when the target person leaves the camera's filming screen, the camera automatically returns to the watching position.
[0045] By adopting the above solution, we can return to the watch position in time to continue tracking other target persons.
[0046] In a second aspect, the present application provides a personnel monitoring system for a power distribution room based on a multi-camera joint tracking algorithm, including: a camera initialization module, used to initialize all monitoring cameras in the power distribution room to a watch position, wherein the watch position is a set horizontal angle and a vertical angle;
[0047] The target person confirmation module is used to use artificial intelligence recognition technology to identify people in the image data collected by all cameras. If a single person is identified, the single person tracking mode is triggered and the single person is determined as the target person;
[0048] The camera adaptive tracking module is used to control the camera closest to the target person to adaptively track when the single-person tracking mode is triggered so that the target person is always in the center of the captured image until the target person leaves the camera capture screen; in the camera adaptive tracking process, the offset of the camera that recognizes the target person is calculated in real time according to the target person's pixels, and the horizontal and vertical viewing angles after the camera is offset are obtained; based on the horizontal and vertical viewing angles after the camera is offset and the world coordinates of the camera, the world coordinates corresponding to the target person's pixels, that is, the world coordinates of the target person, are calculated in real time;
[0049] Switching camera adaptive tracking module; when the target person leaves the shooting screen of the currently controlled camera, switching control to the nearest camera that recognizes the same target person continues to complete adaptive tracking; repeating camera switching control and adaptive tracking of the target person until all cameras fail to recognize the same target person and stop tracking the target person.
[0050] By adopting the above scheme, multiple cameras are controlled to adaptively track and realize real-time tracking and monitoring of the target person, reducing the number of cameras to be installed and reducing costs; the world coordinates of the target person are accurately calculated by using the pixel coordinates of the target person, the camera field of view and the camera world coordinates.
[0051] In a third aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method as described above.
[0052] In a fourth aspect, the present application provides a computer device, comprising a memory, a processor, and a program stored and executable on the memory, wherein the program implements the steps of the above method when executed by the processor.
[0053] In summary, this application has the following beneficial effects:
[0054] 1. Control multiple cameras to adaptively track the identified individual so that the identified individual is always in the center of the shooting screen, reducing the number of cameras while ensuring that the target person is tracked without blind spots; based on the three-dimensional coordinate system, more accurate world coordinates of the target person are obtained through camera field of view, target person pixel coordinates, and camera world coordinate calculation;
[0055] 2. Obtain clearer images of target persons by automatically adjusting the camera magnification, thereby improving the accuracy of target person behavior recognition. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] Figure 1 It is a flow chart of the method for monitoring personnel in a power distribution room based on a multi-camera joint tracking algorithm described in a specific embodiment;
[0057] Figure 2 It is a flowchart of target adaptive tracking in the distribution room personnel monitoring method based on multi-camera joint tracking algorithm described in a specific embodiment;
[0058] Figure 3 It is a structural schematic diagram of the distribution room personnel monitoring system based on the multi-camera joint tracking algorithm described in the specific embodiment. DETAILED DESCRIPTION
[0059] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0060] like Figure 1 As shown, the embodiment of the present application discloses a personnel monitoring system for a power distribution room based on a multi-camera joint tracking algorithm, and the specific steps include:
[0061] S1. Initialize all surveillance cameras in the power distribution room to the watch position.
[0062] Specifically, all monitoring spherical cameras in the power distribution room are marked in the same three-dimensional spatial coordinate system, and each camera is provided with a watch position, which is a certain horizontal angle and vertical angle initially set.
[0063] S2. Use the camera to collect image data of the power distribution room and perform personnel identification.
[0064] Specifically, multiple cameras are used to collect image data of the distribution room, and artificial intelligence recognition technology is used to identify personnel in all currently collected image data of the distribution room, and the identified personnel information is uploaded to the background monitoring platform.
[0065] S3. Determine whether to trigger the single-person tracking mode according to the person recognition result.
[0066] Specifically, if the camera recognizes a single person, the single person tracking mode is triggered, and the recognized single person is determined to be the target person; if the camera recognizes multiple persons, the single person tracking mode is not triggered, and the process goes to S1 to continue collecting the image data of the power distribution room at the next moment and perform personnel identification.
[0067] S4. Control the camera that has identified the target person to start adaptive tracking.
[0068] S41. Determine a control object.
[0069] Specifically, Figure 2 As shown, considering the various situations that may exist when a camera recognizes a target person, such as situation 1: at the same time, there are multiple cameras that recognize the same target person, and camera 1 and camera 2 recognize a single person A. It is only necessary to select a camera that is closest to the target as the control object to complete adaptive tracking; such as situation 2: at the same time, multiple cameras recognize a single person and the recognized single persons are different, such as: camera 1 recognizes a single person A, and camera 2 recognizes a single person B. It is necessary to independently control the camera for each target to complete the adaptive tracking of the specific target. The specific process of determining the control object includes:
[0070] If there are multiple cameras that recognize the target person at the same time, determine whether the recognized target person belongs to the same target person; for multiple cameras that recognize the same target person at the same time, select the camera that recognizes the target person closest to it as the control object, specifically including: calculating the world coordinates of the current target person according to the pixel coordinates of the target person recognized by each camera, calculating the distance between the world coordinates of each camera and the world coordinates of the current target person, and selecting the camera closest to it as the control object; for multiple cameras that recognize different target persons at the same time, for each independent target person, control the camera that is closest to the corresponding target person as the control object.
[0071] S42: Adaptively track the target person according to the determined control object.
[0072] Specifically, when the single-person tracking mode is triggered, the horizontal and vertical angles of the camera that has identified the target person are controlled according to the pixel position of the target person, so that the target person is always in the center of the captured image. In order to better ensure that the target person is always in the center of the captured image, the waist position of the target person is always in the center of the captured image.
[0073] S43, in the process of camera adaptive tracking, obtaining the world coordinates corresponding to the pixels of the target person by real-time calculation.
[0074] Specifically, the offset of the camera that recognizes the target person is calculated in real time according to the target person pixel, and the horizontal and vertical viewing angles after the camera is offset are obtained; the formula is:
[0075]
[0076]
[0077] p1=p+delt_x
[0078] t1=t+delt_y
[0079] Where, delt_x is the offset of the camera horizontal angle, delt_y is the offset of the camera vertical angle; (px, py) is the pixel of the target person; im_w is the width of the image taken by the camera; im_h is the height of the image taken by the camera; fov_w is the horizontal field of view of the camera before the camera offset; fov_h is the vertical field of view of the camera before the camera offset; p1 is the horizontal angle of the camera after the camera offset, p is the horizontal angle of the camera before the camera offset, t1 is the vertical angle of the camera after the camera offset, and t is the vertical angle of the camera before the camera offset;
[0080] Based on the horizontal and vertical viewing angles after the camera is offset, the world coordinates corresponding to the pixels of the target person are calculated in real time according to the world coordinates of the camera, that is, the world coordinates of the target person; the formula is:
[0081] c_p_x=word_x+abs(tan(deg2rad(t1))*word_z) Where,
[0082] c_p_y=word_y
[0083] a_x=
[0084] (c_p_x-word_x)×cos(deg2rad(p1))-(c_p_y-word_y)×sin(deg2rad(p1))+word_x
[0085] a_y=
[0086] (c_p_x-word_x)×sin(deg2rad(p1))+(c_p_y-word_y)×cos(deg2rad(p1))+word_y
[0087] (a_x, a_y, a_z) is the world coordinate of the target person's pixel; (word_x, word_y, word_z) is the world coordinate of the camera.
[0088] S5. Determine whether the target person has left the camera's shooting screen.
[0089] Specifically, during the tracking process, if the target person has not left the camera's shooting screen for adaptive tracking, that is, a single person can still be identified in the camera's shooting screen, then go to step S4 to continue adaptive tracking; if the target person leaves the camera's shooting screen for adaptive tracking, that is, a single person cannot be identified in the camera's shooting screen, then go to S6 and stop controlling the camera's adaptive tracking, and automatically return to the watching position.
[0090] S6. Determine whether there is a camera that has recognized the same target person.
[0091] Specifically, when the target person leaves the currently controlled camera shooting screen, it is determined whether there is a camera that recognizes the same target person; if so, it is transferred to S4 to switch the control camera, that is, the camera closest to the camera that recognizes the same target is selected as the control object among the cameras that recognize the same target, and the tracking is further completed; if not, it is transferred to S7. The same target person refers to the same person and a single person who triggers the single-person tracking mode.
[0092] S7. End the tracking of the target person.
[0093] Through the above method, multiple cameras can be combined to track the same single person, and each camera will follow the single person for adaptive tracking, reducing the number of cameras installed and saving costs; at the same time, in the process of adaptive tracking, the world coordinates of the target person can be accurately obtained, so that the inspection route of the target person can be generated according to the world coordinates of the target person obtained in real time, and the generated inspection route of the target person can be compared with the preset inspection route to check whether there is a missed inspection problem.
[0094] In a specific embodiment, in order to further protect the personal safety of a single person, more accurately identify and timely generate a warning when a single person exhibits a preset behavior, the method further includes:
[0095] After the camera adaptively tracks the target person so that the target person is always in the center of the captured image, the camera magnification is adjusted according to the proportion of the target person in the picture so that the pixel height of the target person is kept within a preset percentage value of the total pixel height of the camera shooting picture. In this embodiment, the preset percentage is set to 80%.
[0096] Artificial intelligence recognition technology is used to identify the behavior of the target person in the picture taken after the camera magnification is adjusted. When a preset behavior is identified, an early warning message is generated and played in real time; the preset behavior includes: climbing behavior, falling behavior, and not wearing a helmet or work clothes correctly. In this embodiment, the artificial intelligence recognition technology uses neural network technology.
[0097] In a specific embodiment, considering that the camera obtains the target person's pixels to calculate the offset and moves with hysteresis, the target person will move in real time, and the camera may not cover the target person after moving according to the calculated offset, resulting in tracking failure. In order to ensure the timeliness of tracking; the method also includes:
[0098] In the process of camera adaptive tracking of the identified target person, the world coordinates of the current target person can be calculated based on the pixel coordinates of the currently identified target person according to the conversion relationship between the world coordinates and the pixel coordinates;
[0099] Calculate the horizontal angle and vertical angle of the camera after it moves when the center of the camera screen tracks the current target person's position, and calculate the camera movement time based on the preset rate of camera movement. The specific formulas for the horizontal angle and vertical angle of the camera include:
[0100] t2=rad2deg(arctan(dis / z2))
[0101]
[0102] In the formula, the calculated world coordinates of the target person are (x1, y1, z1), and the known world coordinates of the camera are (x2, y2, z2);
[0103] The moving speed and direction of the target person are obtained based on the historical image information. Specifically, the xyz coordinates of the target person are calculated through the pixel coordinates of the person in the historical interval 2-3 frames of pictures, and the moving direction and speed of the target person are calculated through the offset of the xyz coordinates and the frame rate.
[0104] According to the moving direction and speed of the target person, the world coordinates of the current target person and the camera movement time are calculated, and the world coordinates of the target person after the camera movement time are estimated.
[0105] When the center of the camera screen tracks the estimated position of the target person after the camera moves, the horizontal angle and vertical angle after the camera moves are calculated, and tracking is performed based on the horizontal angle and vertical angle after the camera moves.
[0106] like Figure 3 As shown, the embodiment of the present application discloses a personnel monitoring system for a power distribution room based on a multi-camera joint tracking algorithm, including:
[0107] The camera initialization module 101 is used to initialize all monitoring cameras in the power distribution room to a watch position, where the watch position is a set horizontal angle and a vertical angle;
[0108] The target person confirmation module 102 is used to use artificial intelligence recognition technology to perform person recognition on the image data collected by all cameras. If a single person is recognized, a single person tracking mode is triggered and the single person is determined as the target person;
[0109] The control camera adaptive tracking module 103 is used to control the camera closest to the target person to adaptively track when the single-person tracking mode is triggered so that the target person is always at the center of the captured image until the target person leaves the captured image of this camera; in the process of camera adaptive tracking, the offset of the camera that recognizes the target person is calculated in real time according to the target person's pixels, and the horizontal and vertical viewing angles after the camera is offset are obtained; based on the horizontal and vertical viewing angles after the camera is offset and the world coordinates of the camera, the world coordinates corresponding to the pixels of the target person, that is, the world coordinates of the target person, are calculated in real time;
[0110] The camera switching adaptive tracking module 104 is used to switch the control distance to the nearest camera that recognizes the same target person to continue adaptive tracking when the target person leaves the currently controlled camera shooting screen; repeat the camera switching control and the adaptive tracking of the target person until all cameras fail to recognize the same target person and stop tracking the target person.
[0111] The system further comprises:
[0112] The target person behavior recognition module 105 is used to use artificial intelligence recognition technology to recognize the target person behavior in the picture taken after the camera magnification is adjusted. When a preset behavior is recognized, an early warning message is generated and played in real time; the preset behavior alarms include: climbing behavior, person falling behavior and improper dressing behavior.
[0113] The camera adaptive tracking control module 104 is also used to control the camera adaptive tracking after identifying the target person so that the target person is always in the center of the captured image, and then complete the camera magnification adjustment according to the proportion of the target person in the picture so that the pixel height of the target person is maintained within a preset percentage value of the total pixel height of the camera shooting picture.
[0114] The embodiment of the present application also discloses a computer-readable storage medium.
[0115] Specifically, the computer-readable storage medium stores a computer program that can be loaded by a processor and executes a method for monitoring personnel in a distribution room based on a multi-camera joint tracking algorithm as described above. The computer-readable storage medium includes, for example: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.
[0116] The embodiment of the present application also discloses a computer device.
[0117] Specifically, the computer device includes a memory and a processor, and the memory stores a computer program that can be loaded by the processor and execute the above-mentioned distribution room personnel monitoring method based on the multi-camera joint tracking algorithm.
[0118] The above are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Any feature disclosed in this specification (including the abstract and drawings), unless otherwise stated, can be replaced by other equivalent or alternative features with similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.
Claims
1. A method for monitoring personnel in a power distribution room based on a multi-camera joint tracking algorithm, characterized in that: include: Initialize all surveillance cameras in the power distribution room to the watch position, where the watch position is the set horizontal angle and vertical angle; Artificial intelligence recognition technology is used to identify people in image data collected by all cameras. If a single person is identified, the single person tracking mode is triggered and the single person is determined as the target person; When the single-person tracking mode is triggered, the camera closest to the target person is controlled to adaptively track so that the target person is always in the center of the captured image until the target person leaves the camera's captured image; during the camera adaptive tracking process, the offset of the camera that recognizes the target person is calculated in real time according to the target person's pixels, and the horizontal and vertical viewing angles after the camera is offset are obtained; based on the horizontal and vertical viewing angles after the camera is offset and the world coordinates of the camera, the world coordinates corresponding to the target person's pixels, i.e., the world coordinates of the target person, are calculated in real time; When the target person leaves the shooting screen of the currently controlled camera, the control is switched to the camera closest to the target person to continue the adaptive tracking; the camera switching control and the adaptive tracking of the target person are repeated until all cameras fail to recognize the target person and stop tracking the target person; The specific calculation formula for calculating the offset of the camera that recognizes the target person in real time according to the target person pixel and calculating the world coordinates corresponding to the target person pixel in real time based on the camera offset and the camera's world coordinates includes: p1=p+delt_x t1=t+delt_y Where, delt_x is the offset of the camera horizontal angle, delt_y is the offset of the camera vertical angle; (px, py) is the pixel of the target person; im_w is the width of the image taken by the camera; im_h is the height of the image taken by the camera; fov_w is the horizontal field of view of the camera before the camera offset; fov_h is the vertical field of view of the camera before the camera offset; p1 is the horizontal angle of the camera after the camera offset, p is the horizontal angle of the camera before the camera offset, t1 is the vertical angle of the camera after the camera offset, and t is the vertical angle of the camera before the camera offset; c_p_x = word_x + abs(tan(deg2rad(t1)) * word_z) where c _ p_y = word_y a_x= (c_p_x-word_x)×cos(deg2rad(p1))-(c_p_y-word_y)×sin(deg2rad(p1))+word_x a_y= (c_p_x-word_x)×sin(deg2rad(p1))+(c_p_y-word_y)×cos(deg2rad(p1))+word_y (a_x, a_y, a_z) is the world coordinate of the target person's pixel; (word_x, word_y, word_z) is the world coordinate of the camera.
2. The method for monitoring personnel in a power distribution room based on a multi-camera joint tracking algorithm according to claim 1 is characterized in that: Also includes: After controlling the camera to adaptively track the target person so that the target person is always in the center of the captured image, the camera magnification is adjusted according to the proportion of the target person in the picture so that the pixel height of the target person is kept within a preset percentage value of the total pixel height of the camera shooting picture; Use artificial intelligence recognition technology to identify the target person's behavior in the picture taken after the camera magnification is adjusted. When the preset behavior is identified, an early warning message is generated and played in real time; The preset behaviors include: climbing, falling, and improper dressing.
3. The method for monitoring personnel in a power distribution room based on a multi-camera joint tracking algorithm according to claim 1 is characterized in that: Also includes: In the process of camera adaptive tracking of the identified target person, the world coordinates of the current target person are calculated based on the pixel coordinates of the currently identified target person; Calculate the horizontal angle and vertical angle of the camera after it moves when the center of the camera screen tracks the current target person, and calculate the camera movement time in combination with the preset rate of camera movement; the specific formulas for the horizontal angle and vertical angle of the camera after it moves are: t2=rad2deg(arctan(dis / z2)) In the formula, the calculated world coordinates of the target person are (x1, y1, z1), and the known world coordinates of the camera are (x2, y2, z2); Obtain the moving speed and direction of the target person based on historical image information, and calculate the current world coordinates of the target person to estimate the world coordinates of the target person after the camera moves; When the center of the camera screen tracks the target person after the camera moves, the horizontal angle and vertical angle of the camera after the camera moves are calculated, and tracking is performed based on the horizontal angle and vertical angle of the camera after the camera moves.
4. The method for monitoring personnel in a power distribution room based on a multi-camera joint tracking algorithm according to claim 1 is characterized in that: The control distance to identify the nearest camera adaptive tracking of the target person includes: If there are multiple cameras that recognize the target person at the same time, determine whether the recognized target person belongs to the same target person; for multiple cameras that recognize the same target person at the same time, calculate the world coordinates of the current target person according to the pixel coordinates of the target person recognized by each camera, calculate the distance between the camera and the target person in combination with the world coordinates of each camera, and select the camera that recognizes the target person closest to it as the control object; For multiple cameras that recognize different target persons at the same time, the camera that recognizes the corresponding target person at the closest control distance is used as the control object for each independent target person.
5. The method for monitoring personnel in a power distribution room based on a multi-camera joint tracking algorithm according to claim 1 is characterized in that: The making the target person always be at the center position of the captured image means that the waist position of the target person is always at the center position of the captured image.
6. The method for monitoring personnel in a power distribution room based on a multi-camera joint tracking algorithm according to claim 1 is characterized in that: When the target person leaves the controlled camera recording screen, the controlled camera automatically returns to the watching position.
7. A personnel monitoring system for a power distribution room based on a multi-camera joint tracking algorithm, characterized in that: include: A camera initialization module is used to initialize all monitoring cameras in the power distribution room to a watch position, where the watch position is a set horizontal angle and vertical angle; The target person confirmation module is used to use artificial intelligence recognition technology to identify people in the image data collected by all cameras. If a single person is identified, the single person tracking mode is triggered and the single person is determined as the target person; The camera adaptive tracking module is used to control the camera closest to the target person to adaptively track when the single-person tracking mode is triggered so that the target person is always in the center of the captured image until the target person leaves the camera capture screen; in the camera adaptive tracking process, the offset of the camera that recognizes the target person is calculated in real time according to the target person's pixels, and the horizontal and vertical viewing angles after the camera is offset are obtained; based on the horizontal and vertical viewing angles after the camera is offset and the world coordinates of the camera, the world coordinates corresponding to the target person's pixels, that is, the world coordinates of the target person, are calculated in real time; Switch the camera adaptive tracking module; When the target person leaves the shooting screen of the currently controlled camera, the control is switched to the camera closest to the target person to continue the adaptive tracking; the camera switching control and the adaptive tracking of the target person are repeated until all cameras fail to recognize the target person and stop tracking the target person; The specific calculation formula for calculating the offset of the camera that recognizes the target person in real time according to the target person pixel and calculating the world coordinates corresponding to the target person pixel in real time based on the camera offset and the camera's world coordinates includes: p1=p+delt_x t1=t+delt_y Where, delt_x is the offset of the camera horizontal angle, delt_y is the offset of the camera vertical angle; (px, py) is the pixel of the target person; im_w is the width of the image taken by the camera; im_h is the height of the image taken by the camera; fov_w is the horizontal field of view of the camera before the camera offset; fov_h is the vertical field of view of the camera before the camera offset; p1 is the horizontal angle of the camera after the camera offset, p is the horizontal angle of the camera before the camera offset, t1 is the vertical angle of the camera after the camera offset, and t is the vertical angle of the camera before the camera offset; c_p_x = word_x + abs(tan(deg2rad(t1)) * word_z) where c _ p_y = word_y a_x= (c_p_x-word_x)×cos(deg2rad(p1))-(c_p_y-word_y)×sin(deg2rad(p1))+word_x a_y= (c_p_x-word_x)×sin(deg2rad(p1))+(c_p_y-word_y)×cos(deg2rad(p1))+word_y (a_x, a_y, a_z) is the world coordinate of the target person's pixel; (word_x, word_y, word_z) is the world coordinate of the camera.
8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 6.
9. A computer device, characterized in that: The computer device comprises a memory, a processor and a program stored and executable on the memory, and the program implements the steps of the method according to any one of claims 1 to 6 when executed by the processor.
