Statistical method for operating personnel in rail-mounted area of rail transit

Abstract

The invention relates to the technical field of rail traffic construction operation rail-mounted area safety, discloses a rail traffic rail-mounted area operating personnel statistical method, and aims to solve the problems of low efficiency, high cost, inaccurate dynamic statistics and large potential safety hazard caused by dependence on manual work or special equipment in rail-mounted area personnel counting. According to the method, a camera, a hard disk video recorder and an edge computing server are installed, a target detection and tracking algorithm is deployed, a virtual area and an entity area are delimited to form a counting area, the number of people entering and exiting is counted and transmitted to a system in real time, the number of people in each area is displayed in real time, and the safety of people in the area is ensured before rail car dispatching. The method does not need operators to wear special equipment, improves the checking efficiency and safety, and is suitable for safety management of rail traffic rail-mounted areas.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of safety technology in the track operation area of rail transit construction, and in particular to a track operation area worker counting method. BACKGROUND

[0002] In the construction and operation and maintenance stage of urban rail transit engineering, the track operation area is the core area of train passing and construction operation, and its safety management is directly related to the life safety of workers and the stable operation of the rail transit system. However, the current counting and control of workers in the track operation area still faces many problems to be solved.

[0003] In the traditional management mode, the personnel entering and leaving the track operation area mainly relies on manual registration, reporting at the way-in and way-out points or handheld terminal statistics, which is not only low in efficiency, but also easy to cause data omission or errors due to human negligence. For example, during the peak construction period, when multiple teams cross operation, manual counting is difficult to synchronize the personnel dynamics in each area in real time, and if the actual number of personnel in the operation area cannot be accurately mastered before the track car is dispatched, it is easy to cause safety accidents such as collision.

[0004] In order to improve the management accuracy, some scenes try to introduce a personnel positioning system, which realizes location tracking by combining base stations through special equipment (such as badges, bracelets, etc.) with positioning chips worn by workers. However, this solution has significant limitations: on the one hand, the deployment cost of positioning chips and base stations is high, especially in long-distance track operation areas (such as subway tunnels and ground tracks), and the device maintenance and signal coverage are extremely difficult; on the other hand, workers often cause positioning failure due to inconvenient equipment wearing, insufficient battery life or operational negligence, and it is difficult to ensure coverage in actual management, which still has safety risks.

[0005] In addition, the track operation area is usually a narrow space, and there are complex environments such as curves, tunnels, alternating light and dark, etc. The traditional monitoring system can only realize video recording and cannot be directly used for personnel quantity statistics and regional dynamic management. At the same time, the division of operation areas in different construction stages is flexible and variable, and the fixed physical division statistics method cannot adapt to the dynamic adjustment demand, so that the management personnel cannot quickly obtain the real-time personnel data of a specific area, affecting the efficiency of track car dispatching and construction progress.

[0006] In summary, the existing technology has problems such as low efficiency, high cost, poor adaptability and prominent safety risks in the personnel counting in the track operation area, and there is an urgent need for a personnel counting method that does not rely on special equipment, can flexibly adapt to regional division, and has real-time and accuracy, to meet the actual needs of safety management in the track operation area of rail transit. SUMMARY

[0007] The present application intends to provide a rail transit track area operation personnel statistical method to solve the problem that current rail transit track area operation personnel counting relies on manual or special equipment, which has low efficiency, high cost, inaccurate dynamic statistics and great safety hazards.

[0008] To achieve the above-mentioned purpose, the present application adopts the following technical solutions: A rail transit track area operation personnel statistical method comprises: S1, installing and deploying a camera, a hard disk recorder and an edge computing server in the track area, the camera is connected to the hard disk recorder through a network, the hard disk recorder is connected to the edge computing server, and the edge computing server is connected to a track area personnel counting system server through a network; S2, deploying a target detection and target tracking algorithm in the edge computing server, the target detection algorithm is used to identify personnel in the monitoring range of the camera, and the target tracking algorithm is used to determine the walking direction of the personnel and determine whether the personnel enter or leave the delineated area; S3, delineating a virtual area in each camera area, and the virtual area and the entity area jointly constitute a personnel counting area; S4, the edge computing server counts the number of people entering and leaving each camera based on the target detection and target tracking algorithm, and transmits the data to the rail transit track area operation personnel counting system in real time; S5, according to the area delineated by the camera, real-time display of the real-time number of people in each area is realized; S6, before the rail car is dispatched, the number of personnel in the travel area is checked, and after it is ensured that there is no personnel, the rail car is dispatched to the area.

[0009] The principle and advantages of the scheme are: in actual application, the camera is connected to the hard disk recorder through the network, the hard disk recorder is connected to the edge computing server, and each edge computing server is connected to the system server through the network to form a data transmission link to ensure smooth transmission of information between devices. The total number of personnel in the current camera picture is identified through target detection, but the total number of personnel is instantaneous and static, and it cannot be determined whether personnel are replaced, personnel flow across the lens area, and personnel visual obstruction, etc. Therefore, by target tracking, each personnel is assigned a unique identifier and their movement trajectory is tracked to accurately determine entry or exit behavior. Target detection serves as the basis for identifying personnel, and target tracking serves as a necessary means to determine personnel flow direction and achieve dynamic statistics of the number of personnel remaining in the area. The combination of the two achieves the real-time, accurate, and dynamic personnel counting needs of the track area. The track area is mostly narrow and long, and the work area is dynamically adjusted according to the construction progress. Virtual area division can adapt to the complex scene of the track area, is not limited by physical space, accurately counts the number of personnel in a specific area, and combines virtual areas with physical areas. The combination of virtual areas and physical areas not only takes advantage of the flexibility of virtual areas, but also relies on the physical certainty of physical areas, making personnel counting both accurate and dynamic, and firmly anchoring the physical boundaries of safety management, providing more reliable technical support for track area safety management. The scheme can count the personnel in the track area without the need for personnel to wear special equipment, and can serve safety management supervision while counting the number of personnel.

[0010] Preferably, as an improvement, in S2, the target detection algorithm identifies personnel in the camera monitoring range including: S21, extracting continuous image frames from the video stream collected by the camera at a preset frame rate; S22, extracting features from the image frames through the YOLOv5 model to identify all personnel targets in the frame, and outputting the bounding box coordinates and confidence of the personnel targets; S23, filtering the detection results with a confidence lower than a preset threshold to retain valid personnel targets; S24, real-time output of the position information of the valid personnel targets to the target tracking algorithm.

[0011] Technical effect: The deep learning model accurately identifies personnel targets in various postures in the track area, and the confidence filtering mechanism reduces misidentification, providing high-quality target data input for the subsequent tracking algorithm, ensuring the accuracy and stability of personnel identification.

[0012] Preferably, as an improvement, in S2, the target tracking algorithm implements personnel walking direction determination and area entry and exit identification including: S25, receiving the personnel target bounding box coordinates output by the target detection algorithm and assigning a unique tracking ID to each personnel; S26, the Kalman filter algorithm is used to predict the motion trajectory of each tracking ID, and the trajectory parameters are updated in combination with the current frame detection result; S27, the intersection of the motion trajectory of the tracking ID and the boundary of the virtual area is calculated in real time, and the motion vector at the intersection is extracted; S28, the walking direction of the personnel is determined according to the direction of the motion vector: when the vector points to the inside of the virtual area, it is marked as entering and accumulated; when the vector points to the outside of the virtual area, it is marked as leaving and accumulated.

[0013] Technical effect: through the unique ID association, the personnel continuous tracking is realized, the trajectory prediction and the boundary intersection analysis are combined, the behavior of the personnel entering and leaving the area is accurately determined, the problem that the personnel flow direction cannot be distinguished by simply target detection is solved, and dynamic data support is provided for the area resident number statistics.

[0014] Preferably, as an improvement, in S28, the calculation of the motion vector is based on the fitting of more than 3 consecutive trajectory points, the motion direction vector is determined by the least square method, and the direction determination error is not more than 5 degrees.

[0015] Technical effect: through multi-frame trajectory fitting and mathematical optimization algorithm, the interference of single-frame image noise on direction determination is reduced, the accuracy of personnel walking direction recognition is ensured, and the number statistics deviation caused by direction misjudgment is reduced.

[0016] Preferably, as an improvement, it further includes S29, when the personnel target moves across the camera, the tracking ID of the adjacent camera is matched through the clothing and body shape features, and when the matching is consistent, the tracking ID is unified.

[0017] Technical effect: realize the trajectory connection across devices, and ensure the continuity of the entry and exit counting.

[0018] Preferably, as an improvement, in S29, the cross-camera feature matching adopts ORB feature point extraction and FLANN matching algorithm, and the matching similarity threshold of the same personnel ID is set to not less than 0.75.

[0019] Technical effect: it is convenient to ensure the accuracy of ID association.

[0020] Preferably, as an improvement, S3 includes: S31, through the graphical interactive interface provided by the personnel counting system in the track area, load the real-time monitoring picture or historical picture frame of each camera as the bottom picture; S32, the boundary of the virtual area is manually outlined on the bottom picture through the polygon drawing tool, or the template area associated with the mileage post number and device coordinates in the track area is selected as the virtual area; S33, the system maps the outlined virtual area boundary coordinates with the corresponding camera's physical parameters, generating the spatial coordinate boundary within the camera's field of view; S34, the spatial position relationship between the virtual area and the entity area is associated, and the physical boundary coordinates of the entity area are synchronously converted into coordinates within the camera's field of view, which together with the virtual area boundary form a complete electronic fence of the personnel counting area.

[0021] Technical effect: Through graphical interaction and coordinate mapping technology, flexible demarcation of virtual area and accurate association of entity area are realized, ensuring the consistency of electronic fence and actual space of the track area, and providing accurate boundary reference for personnel access determination.

[0022] Preferably, as an improvement, the demarcation of the virtual area adopts GIS map-based coordinate mapping technology, which calibrates the three-dimensional coordinate system of the track area line with the imaging coordinate system of the camera, so that the virtual area boundary is directly associated with the physical identifier of the track area, including but not limited to actual mileage and equipment position.

[0023] Technical effect: accurate mapping of virtual area and track area physical space is realized, so that the management personnel can intuitively understand the virtual area range through familiar physical identifiers (such as mileage post and equipment position), improving the convenience and accuracy of area division.

[0024] Preferably, as an improvement, the virtual area supports timing automatic adjustment, which automatically loads the virtual area boundary of the corresponding work range at the specified time period through the system preset work plan time parameter, realizing dynamic adaptation to the construction progress.

[0025] Technical effect: the virtual area can be automatically adjusted according to the work plan, without the need for manual repeated operation, adapting to the dynamic change of the work range of the track area, and improving the adaptability and management efficiency of the system to the construction progress.

[0026] Preferably, as an improvement, the combination of the virtual area and the entity area adopts boundary fusion technology, when the physical boundary of the entity area appears in the camera monitoring picture, the system automatically identifies the physical boundary and takes it as a supplementary boundary of the virtual area, ensuring that the counting area has no monitoring blind area.

[0027] Technical effect: through the supplement of the virtual area by the entity boundary, the monitoring blind area caused by the omission of virtual area demarcation or the limitation of camera view angle is eliminated, ensuring that the personnel counting covers all physical spaces that need to be controlled, and improving the comprehensiveness of safety management. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 A structural schematic diagram of a track traffic track area work personnel counting method. DETAILED DESCRIPTION

[0029] Further details are described below by way of specific embodiments: The embodiments are substantially as shown in the accompanying Figure 1 drawings: A track traffic track area operation personnel statistical method, comprising: S1, installing and deploying a camera, a hard disk video recorder and an edge computing server in the track area, the camera is connected with the hard disk video recorder through a network, the hard disk video recorder is connected to the edge computing server, and the edge computing server is connected to a track area personnel counting system server through a network.

[0030] Specifically, in the embodiment, a high-definition network camera with a wide dynamic range (≥120 dB) and infrared night vision function is adopted, the resolution is not less than 2 million pixels, the frame rate supports 25 fps / 30 fps adjustable, and an electric zoom lens is selected according to the monitoring distance; the installation position is arranged at intervals along the track area line direction, the camera in the tunnel is fixed to the side wall at a preset height from the track surface, and the lens is directed downward at an angle with the track; the ground section camera is installed on the overhead contact system column or a special support, ensuring that the monitoring picture covers the track center to each side 3 meters range, and the overlapping rate of adjacent camera monitoring ranges is not less than 15%. An industrial-grade embedded hard disk video recorder is adopted, which supports at least 16-way high-definition video input, is equipped with several enterprise-level monitoring hard disks, supports RAID5 redundant backup, has a continuous video recording capability of 7x24 hours and a video storage capacity of not less than 30 days; the hard disk video recorder is installed in the signal room or interval substation along the track area, redundant power supply is realized through double power modules to avoid video interruption caused by single point failure. A compact industrial edge server is adopted, which supports GPU acceleration and is deployed in the same machine room with the hard disk video recorder, is connected with the hard disk video recorder through a gigabit industrial Ethernet switch, a single edge computing server is responsible for managing algorithm processing tasks of not more than 32 cameras, and supports cooperative calculation between edge nodes. The track area personnel counting system server is deployed in the operation control center machine room.

[0031] S2, deploying a target detection algorithm and a target tracking algorithm in the edge computing server, the target detection algorithm is used to identify personnel in the camera monitoring range; the target tracking algorithm is used to determine the walking direction of the personnel and determine whether the personnel enter or leave the delineated area. The target detection alone cannot achieve dynamic statistics of the number of personnel remaining in a specific area. The target detection can identify the personnel present in the current image and count them, but this count is instantaneous and static, and only the number of people visible to the camera at the moment can be obtained. It cannot distinguish between original personnel in the area and new personnel entering / leaving the area. For example, there are initially 3 people in a virtual area, and 10 minutes later, the target detection still shows 3 people, but in fact, one person has left and one person has entered, resulting in a replacement. Relying solely on target detection will misjudge that there is no flow of people, while the actual number of people remaining has not changed, but the personnel have been replaced. This dynamic change is crucial to the safety management of the track area (such as confirming whether unregistered personnel have entered). The track area usually needs to be covered by multiple cameras (such as segmented monitoring of long tunnels). The target detection of a single camera can only count the number of people in its field of view, and cannot track the movement of personnel from the monitoring range of camera A to the monitoring range of camera B. For example, personnel walk from area 1 (covered by camera A) to area 2 (covered by camera B). Using only target detection will result in a decrease of 1 person in area 1 and an increase of 1 person in area 2, and the statistical results of both areas will be inaccurate. In the track area, there are obstructions (such as equipment, tunnel pillars) or personnel temporarily walking out of the camera's field of view (such as bending down to work below the track). Target detection will misjudge that the personnel have left due to temporary invisibility, resulting in a sudden decrease in the count. Therefore, by combining target detection and target tracking, the statistical accuracy can be improved.

[0032] The target detection algorithm identifies personnel in the camera monitoring range, including: S21, extracting continuous image frames from the video stream collected by the camera at a preset frame rate. In this embodiment, the preset frame rate is not less than 25 frames per second.

[0033] S22, extracting features from the image frames through a YOLOv5 model to identify all personnel targets in the frames, and outputting the bounding box coordinates and confidence of the personnel targets. The target detection algorithm also excludes features of non-personnel targets in the track area, including at least railcars, large maintenance equipment in the track area, and static infrastructure.

[0034] S23, filtering the detection results with a confidence lower than a preset threshold. In this embodiment, the preset threshold for confidence is 0.8, and valid personnel targets are retained.

[0035] S24, real-time outputting the position information of the valid personnel targets to the target tracking algorithm.

[0036] The target tracking algorithm determines the walking direction of the personnel and identifies whether the personnel enter or leave the area, including: S25, receive the personnel target bounding box coordinates output by the target detection algorithm, and assign a unique tracking ID to each personnel.

[0037] S26, use the Kalman filter algorithm to predict the motion trajectory of each tracking ID, and update the trajectory parameters combined with the current frame detection results. For scenes with occlusion lasting no more than 1 second, maintain trajectory continuity.

[0038] S27, calculate the intersection of the motion trajectory of the tracking ID and the boundary of the virtual area in real time, and extract the motion vector at the intersection.

[0039] S28, determine the walking direction of the personnel according to the direction of the motion vector: when the vector points to the inside of the virtual area, mark it as entering and accumulate Snj; when the vector points to the outside of the virtual area, mark it as leaving and accumulate Snc. The calculation of the motion vector is based on the fitting of more than 3 consecutive trajectory points, and the motion direction vector is determined by the least square method, with a direction determination error of not more than 5 degrees.

[0040] Also includes S29, when the personnel target moves across the camera, match the tracking ID of the adjacent camera through clothing and body shape features, and when the match is consistent, unify the tracking ID. Cross-camera feature matching uses ORB feature point extraction and FLANN matching algorithm, and the matching similarity threshold of the same personnel ID is set to not less than 0.75, ensuring the accuracy of ID association.

[0041] S3, delimit a virtual area in each camera area, and the virtual area and the entity area together form a personnel counting area, which can form a complement to each other by combining the characteristics of the two, significantly improving the accuracy, flexibility and adaptability of the personnel counting in the walking area.

[0042] The virtual area is flexibly defined according to the work requirements, such as freely adjusting the range along the extension direction of the track to adapt to the work range of the track area dynamically changing with the construction progress; the entity area (such as the physical boundary of the entity fence, the tunnel wall, etc.) provides rigid constraints to avoid the virtual area from being too flexible and being out of touch with the actual safety boundary. The entity area may have monitoring dead angles (such as the corner of the fence, the equipment shielding place), and the virtual area can flexibly cover these blind areas based on the camera field of view, and realize no-dead-angle monitoring through algorithm linkage; at the same time, the boundary of the virtual area can be associated with the physical identifier (such as the column, the turnout) of the entity area, to ensure that the behavior of personnel entering and leaving the entity boundary is accurately captured by the algorithm of the virtual area, avoiding the incomplete area division leading to missed statistics. The virtual area can also refine the statistical unit (such as the tunnel maintenance area, the track laying area) according to the work team, the task type, etc., to meet the fine management requirements; the entity area provides a macro physical partition framework (such as the up-line section, the down-line section), and the combination of the two can not only display the real-time number of each sub-area, but also realize global summary statistics through the entity area framework, giving consideration to local control and overall scheduling. The track car scheduling needs to be based on the premise of no personnel, and the physical boundary of the entity area is the basic red line for safety scheduling, while the virtual area further divides the pre-warning area, the core area and other sub-ranges in the entity area, and when the personnel approach the entity boundary, the real-time statistics of the virtual area can trigger the early warning; the check-in area composed of the two ensures that the no-personnel state of the pre-checking covers all the space within the physical boundary, avoiding the safety loophole caused by single area division.

[0043] Specifically, S3 further includes: S31, loading the real-time monitoring picture or the historical picture frame of each camera as a bottom map through the graphical interactive interface provided by the track area personnel check-in system.

[0044] S32, manually outlining the boundary of the virtual area on the bottom map through a polygon drawing tool, or selecting a template area associated with the track area mileage post number and equipment coordinates as the virtual area.

[0045] S33, the system maps and converts the outlined boundary coordinates of the virtual area and the physical parameters of the corresponding camera to generate the spatial coordinate boundary within the field of view of the camera. The virtual area is defined by using the coordinate mapping technology based on the GIS map, which calibrates the three-dimensional coordinate system of the track area line and the imaging coordinate system of the camera, so that the boundary of the virtual area is directly associated with the physical identifier of the track area, including but not limited to the actual mileage and equipment position.

[0046] S34, the spatial position relationship of the virtual area and the entity area is associated, the physical boundary coordinates of the entity area are synchronously converted into the coordinates in the camera field of view, and the virtual area boundary is collectively composed to form a complete electronic fence of the personnel counting area. The combination of the virtual area and the entity area adopts a boundary fusion technology. When the physical boundary of the entity area appears in the camera monitoring picture, the system automatically identifies the physical boundary and takes it as a supplementary boundary of the virtual area, so that no monitoring blind area is ensured in the counting area. The virtual area supports timing automatic adjustment. Through the preset work plan time parameter of the system, the virtual area boundary of the corresponding work range is automatically loaded in the specified time period, so that dynamic adaptation to the construction progress is realized.

[0047] S4, the edge computing server counts the number of people entering and leaving each camera based on the target detection and target tracking algorithm, and transmits the data to the rail transit track area operation personnel counting system in real time. Specifically, the number of people in the set area is R, specifically R1, R2,..., Rn; the camera is named S, specifically S1, S2,..., Sn; in the irradiation area of the set camera, the person facing the camera lens is defined as the entering number, and is named as Snj; the person facing away from the camera lens is defined as the leaving number, and is named as Snc; the edge computing server counts the number of people entering and leaving each camera Snj and Snc based on the target detection and target tracking algorithm, and transmits the data to the rail transit track area operation personnel counting system in real time; the number of operation personnel in each area is Rn, Rn=Snj-Snc value of all cameras in the virtual area, plus Snc-Snj value of a plurality of cameras outside the adjacent virtual area constituting the virtual area boundary.

[0048] S5, according to the area delineated by the camera, the real-time number of people in each area is displayed in real time; a plurality of areas can be selected continuously, and the display is summarized; the entire track area operation personnel R=R1+R2+……+Rn.

[0049] S6, before the rail car is dispatched, the number of personnel in the travel area is checked, and after it is ensured that there is no personnel, the rail car is dispatched to the area.

[0050] The above only describes the embodiments of the present application, and the specific technical solutions and / or common knowledge of the scheme are not described in detail. It should be noted that for those skilled in the art, without departing from the technical solutions of the present application, a plurality of deformations and improvements can be made, which should also be considered as the protection scope of the present application, which will not affect the effect and practicality of the patent. The protection scope claimed in the present application should be subject to the content of its claims, and the specific embodiments in the description can be used to explain the content of the claims.

Claims

1. A method for statistical analysis of personnel working in rail transit track areas, characterized in that, include: S1, cameras, hard disk recorders and edge computing servers are installed and deployed in the track area. The cameras and hard disk recorders are connected to the network, the hard disk recorders are connected to the edge computing server, and the edge computing server is connected to the track area personnel counting system server through the network. S2 deploys target detection and target tracking algorithms in the edge computing server. The target detection algorithm is used to identify people within the camera's monitoring range; the target tracking algorithm is used to determine the direction of people's movement and to determine whether people are entering or leaving the designated area. S3 defines a virtual area within each camera area, and the virtual area and the physical area together form the personnel counting area; S4, the edge computing server counts the number of people entering and exiting each camera based on target detection and target tracking algorithms, and transmits the data to the rail transit track area staff counting system in real time; S5 displays real-time population information for each area defined by the camera. S6. Before dispatching a railcar, check the number of people in the area to be traveled. Once you are sure there are no people, dispatch the railcar to the area.

2. The method for statistical analysis of personnel working in rail transit track areas according to claim 1, characterized in that, In step S2, the target detection algorithm identifies the following personnel within the camera's monitoring range: S21, extract continuous image frames from the video stream captured by the camera at a preset frame rate; S22: The YOLOv5 model is used to extract features from the image frame, identify all human targets within the frame, and output the bounding box coordinates and confidence scores of the human targets. S23, Filter out detection results with confidence levels below a preset threshold, retaining valid personnel targets; S24 outputs the location information of the effective personnel target to the target tracking algorithm in real time.

3. The method for statistical analysis of personnel working in rail transit track areas according to claim 2, characterized in that, In step S2, the target tracking algorithm for determining the direction of movement of people and identifying areas of entry and exit includes: S25, receive the bounding box coordinates of the personnel target output by the target detection algorithm, and assign a unique tracking ID to each person; S26, use the Kalman filter algorithm to predict the motion trajectory of each tracking ID, and update the trajectory parameters by combining the detection results of the current frame; S27, calculate the intersection of the motion trajectory of the tracking ID and the boundary of the virtual region in real time, and extract the motion vector at the intersection point; S28. Determine the direction of movement of a person based on the direction of the motion vector: when the vector points into the virtual area, mark it as entering and accumulate; when the vector points out of the virtual area, mark it as leaving and accumulate.

4. The method for statistical analysis of personnel working in rail transit track areas according to claim 3, characterized in that: In step S28, the motion vector is calculated based on the fitting of trajectory points in more than 3 consecutive frames, and the motion direction vector is determined by the least squares method, with the direction determination error not exceeding 5 degrees.

5. The method for statistical analysis of personnel working in rail transit track areas according to claim 3, characterized in that: It also includes S29, which matches the tracking IDs of adjacent cameras with clothing and body features when a person moves across cameras. If the matches are consistent, the tracking IDs are unified.

6. The method for statistical analysis of personnel working in rail transit track areas according to claim 5, characterized in that: In step S29, cross-camera feature matching uses ORB feature point extraction and FLANN matching algorithm, and the matching similarity threshold for the same person ID is set to be no less than 0.

75.

7. The method for statistical analysis of personnel working in rail transit track areas according to claim 1, characterized in that, S3 includes: S31 uses the graphical interface provided by the personnel count system in the track area to load real-time monitoring images or historical frames from each camera as the base image. S32. On the base map, manually outline the boundaries of the virtual area using the polygon drawing tool, or select a template area preset by the system that is associated with the track area mileage station and equipment coordinates as the virtual area. S33, the system maps and transforms the coordinates of the outlined virtual area boundary with the physical parameters of the corresponding camera to generate the spatial coordinate boundary within the field of view of the camera. S34 associates the spatial relationship between the virtual area and the physical area, synchronously converts the physical boundary coordinates of the physical area into coordinates within the camera's field of view, and together with the boundary of the virtual area, forms a complete electronic fence for the personnel counting area.

8. A method for statistical analysis of personnel working in rail transit track areas according to claim 7, characterized in that: The virtual area is delineated using GIS map-based coordinate mapping technology. By calibrating the three-dimensional coordinate system of the track area with the imaging coordinate system of the camera, the boundary of the virtual area is directly associated with the physical identifiers of the track area. These physical identifiers include, but are not limited to, actual mileage and equipment location.

9. A method for statistical analysis of personnel working in rail transit track areas according to claim 7, characterized in that: The virtual area supports automatic adjustment at set times. By using the system's preset work plan time parameters, the virtual area boundary of the corresponding work area is automatically loaded during a specified time period to achieve dynamic adaptation with the construction progress.

10. A method for statistical analysis of personnel working in rail transit track areas according to claim 7, characterized in that: The combination of the virtual area and the physical area adopts boundary fusion technology. When the physical boundary of the physical area appears in the camera monitoring screen, the system automatically identifies the physical boundary and uses it as a supplementary boundary of the virtual area to ensure that there are no blind spots in the inventory area.

Images