A tunnel personnel positioning method based on video recognition

Abstract

The application provides a tunnel personnel positioning method based on video recognition, comprising the following steps: arranging a base pile and a video monitoring device, determining the visual range of the video monitoring device; drawing a virtual paragraph mark, realizing the determination of the position of the video monitoring area corresponding to the video monitoring image; establishing the position of the virtual grid in the video monitoring image, and mapping the actual position of the tunnel physical space; and positioning personnel. The application provides a tunnel personnel positioning method based on video recognition, which does not require personnel to wear a positioning device, can intuitively and accurately and quickly determine the position and trajectory of personnel through video monitoring image information, reduces the interference of various complex environments in the tunnel on the positioning device, and has the advantages of high positioning efficiency and accurate positioning result.

Description

Technical Field

[0001] This invention belongs to the field of tunnel personnel positioning technology, specifically relating to a tunnel personnel positioning method based on video recognition. Background Technology

[0002] Tunnel construction is characterized by harsh conditions, severe environment, high risk, and a large number of personnel involved, which poses significant challenges to supervision and safety. Therefore, it is necessary to conduct efficient and accurate location tracking of personnel inside the tunnel.

[0003] Currently, the commonly used method for personnel positioning in tunnels is to use wireless pulse technology to locate personnel. This involves deploying positioning base stations within the tunnel and combining this with positioning tags worn by construction workers, such as safety helmets and electronic work badges. However, this existing wireless pulse technology-based positioning method has the following problems: 1) It cannot directly reflect the personnel's status: Current methods only utilize wireless pulse signals. If a positioning tag remains stationary in the system for an extended period, due to the lack of on-site images, it is impossible to determine whether the stationary status is due to a lost tag or the stationary status of the personnel wearing the tag. 2) The complex environment inside the tunnel, coupled with interference from various mechanical devices, can cause inaccurate positioning tag signals. Summary of the Invention

[0004] In view of the shortcomings of existing technologies, the present invention provides a tunnel personnel positioning method based on video recognition, which can effectively solve the above problems.

[0005] The technical solution adopted in this invention is as follows:

[0006] This invention provides a method for locating personnel in tunnels based on video recognition, comprising the following steps:

[0007] Step 1: Install foundation piles and video surveillance equipment.

[0008] Along the longitudinal direction of the tunnel, on one side of the tunnel construction area, multiple foundation piles are arranged at intervals, each foundation pile having a foundation pile number related to the mileage; on the other side of the tunnel construction area, multiple video monitoring devices are arranged at intervals.

[0009] The spacing between the video surveillance devices is equal to the visible range of the video surveillance devices along the longitudinal direction of the tunnel, which is S meters.

[0010] Step 2: When each video monitoring device captures video images, its shooting angle is directly facing the tunnel surface and remains stationary. Its video field of view covers its corresponding video monitoring area. The length of the video monitoring area is equal to the visible range of the video monitoring device, which is S meters; the width of the video monitoring area is equal to the width of the tunnel surface, which is W meters. This ensures the integrity and continuity of the video monitoring images captured by each video monitoring device, and can completely cover the road surface of the tunnel construction area.

[0011] Step 3: For each video surveillance image captured by a video surveillance device, pre-draw virtual segment markers to determine the location of the video surveillance area corresponding to the video surveillance image;

[0012] Specifically, between any two adjacent foundation piles, multiple virtual pile numbers are generated at intervals of S meters within the visible range, and the name of each virtual pile number is related to the mileage.

[0013] For any video surveillance image captured by a video surveillance device, its starting and ending points along the length direction are marked using the foundation pile numbers and / or virtual pile numbers at both ends of the corresponding video surveillance area. The foundation pile numbers and / or virtual pile numbers are superimposed on the starting and ending points of the video surveillance image to form virtual segment marks of the video surveillance image. Through the virtual segment marks, the actual mileage position of the starting and ending points of the video surveillance image within the tunnel can be directly converted.

[0014] Step 4: Perform video virtual meshing processing on the video surveillance images captured by each video surveillance device to form multiple virtual meshes, and establish the mapping between the position of the virtual mesh in the video surveillance image and the actual position in the physical space of the tunnel.

[0015] For video surveillance images with overlaid virtual segment markers, the images are gridded according to the set virtual grid size to form a virtual grid with a columns and b rows, where a is the number of columns and b is the number of rows. For the virtual grid in the i-th column and j-th row of the video surveillance image, where i∈[1,a] and j∈[1,b], the following algorithm is used to obtain its actual position (x0,y0) in the tunnel physical space, where x0 and y0 are the horizontal and vertical coordinates of its actual position in the tunnel physical space, respectively.

[0016] In the video surveillance image, a rectangular coordinate system is established with the length direction (the longitudinal direction of the tunnel) as the y-axis, the width direction (the transverse direction of the tunnel) as the x-axis, and the lower left corner of the video surveillance image as the origin.

[0017] The virtual segment marker corresponding to the origin of the coordinate system is the foundation pile number and / or virtual pile number marker of the starting point of the video surveillance image. Using this virtual segment marker, the actual mileage position of the origin of the coordinate system within the tunnel is obtained, denoted as M. The actual position (x0, y0) is obtained using the following formula:

[0018] x0=j*g

[0019] y0=M+i*g

[0020] Where: g is the side length of the virtual grid;

[0021] Step 5: When personnel location is required, obtain the video surveillance images captured by each video surveillance device; then, use a target recognition algorithm to identify personnel in each video surveillance image, assuming that a person is identified in video surveillance image P.

[0022] Then, the virtual grid where the person is located is first determined in the video surveillance image P, and the position of the virtual grid is mapped to the actual position in the physical space of the tunnel, thereby achieving accurate positioning of the person.

[0023] Preferably, in step 1, each foundation pile has a foundation pile number related to its mileage, specifically:

[0024] Along the longitudinal direction of the tunnel, between the starting and ending points of the tunnel construction area, on one side of the tunnel, a total of n foundation piles are arranged, and each foundation pile is numbered as follows: pile number K0, pile number K1, ..., pile number K(n-2), pile number K(n-2)+L2; among them, the foundation piles corresponding to pile number K0, pile number K1, ..., pile number K(n-2) are arranged at equal intervals with an interval of L1 meters; the foundation pile of pile number K(n-2)+L2 represents a distance of L2 meters from the foundation pile of the adjacent pile of pile number K(n-2) in front of it, where L2≤L1.

[0025] Preferably, L1 meter and L2 meter are both integer multiples of the visible range S meter of the video surveillance device.

[0026] Preferably, for the first video surveillance area between chainage K0 and chainage K1, the second video surveillance area between chainage K1 and chainage K2, ..., the (n-2)th video surveillance area between chainage K(n-3) and chainage K(n-2), L1 / S video surveillance devices are arranged at equal intervals.

[0027] For the (n-1)th video surveillance area between chainage K(n-2) and chainage K(n-2)+L2, L2 / S video surveillance devices are arranged at equal intervals.

[0028] Preferably, for the first video surveillance area between station K0 and station K1, a total of L1 / S virtual station numbers are generated in sequence, namely: station K0+S, station K0+2S, ..., station K0+(L1 / S)*S=K0+L1;

[0029] Therefore, among the L1 / S video surveillance devices arranged in the first video surveillance area, the starting point and ending point markers of the first video surveillance device are respectively: station number K0, station number K0+S; the starting point and ending point markers of the second video surveillance device are respectively: station number K0+S, station number K0+2S; and so on, the starting point and ending point markers of the L1 / S video surveillance devices are respectively: station number K0+L1, station number K1.

[0030] Preferably, when tracking the location and trajectory of personnel, the specific steps include:

[0031] The video surveillance images captured by each video surveillance device at the current moment are used to form a video surveillance image sequence; a target recognition algorithm is used to detect the location of personnel.

[0032] Then, using a target tracking algorithm, the position of the same person in various video surveillance devices at the next moment is continuously tracked to achieve personnel trajectory tracking.

[0033] The tunnel personnel positioning method based on video recognition provided by this invention has the following advantages:

[0034] This invention proposes a tunnel personnel positioning method based on video recognition. It eliminates the need for personnel to wear positioning devices and can accurately and quickly determine the location and trajectory of personnel by intuitively and quickly using video monitoring image information. It also reduces the interference of various complex environments in the tunnel on the positioning equipment and has the advantages of high positioning efficiency and accurate positioning results. Attached Figure Description

[0035] Figure 1 A layout diagram for a tunnel personnel positioning method based on video recognition provided by the present invention;

[0036] Figure 2 This is a schematic diagram illustrating the principle of mapping the virtual grid position to the actual physical position of the tunnel, as provided by this invention. Detailed Implementation

[0037] To make the technical problems solved, the technical solutions, and the beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the invention.

[0038] This invention belongs to the field of image recognition, specifically relating to the application of an image algorithm based on video surveillance in tunnel personnel positioning. This invention proposes a tunnel personnel positioning method based on video recognition, which eliminates the need for personnel to wear positioning devices. It can intuitively and accurately determine the location and trajectory of personnel through video surveillance image information, while also reducing the interference of various complex environments in the tunnel on the positioning device. It has the advantages of high positioning efficiency and accurate positioning results.

[0039] This invention uses common video surveillance methods for personnel location. Video surveillance, as the most commonly used IoT monitoring device, has the advantages of low cost, stability, intuitiveness, and minimal susceptibility to electromagnetic interference.

[0040] This invention provides a method for locating personnel in tunnels based on video recognition, with reference to... Figure 1 This includes the following steps:

[0041] Step 1: Install foundation piles and video surveillance equipment.

[0042] Along the longitudinal direction of the tunnel, on one side of the tunnel construction area, multiple foundation piles are arranged at intervals, each foundation pile having a foundation pile number related to the mileage; on the other side of the tunnel construction area, multiple video monitoring devices are arranged at intervals.

[0043] The spacing between the video surveillance devices is equal to the visible range of the video surveillance devices along the longitudinal direction of the tunnel, which is S meters.

[0044] In this step, as a specific implementation method, each foundation pile has a foundation pile number related to its mileage, specifically:

[0045] Along the longitudinal direction of the tunnel, between the starting and ending points of the tunnel construction area, on one side of the tunnel, a total of n foundation piles are arranged and numbered as follows: pile number K0, pile number K1, ..., pile number K(n-2), pile number K(n-2)+L2; where the foundation piles corresponding to pile number K0, pile number K1, ..., pile number K(n-2) are arranged at equal intervals with an interval of L1 meters; the foundation pile at pile number K(n-2)+L2 represents a distance of L2 meters from the foundation pile at pile number K(n-2) in front of it, where L2≤L1, and L1 meters and L2 meters are both integer multiples of the visible range S meters of the video monitoring device.

[0046] As an example, the video surveillance device is arranged in the following manner:

[0047] For the first video surveillance area between chainage K0 and chainage K1, the second video surveillance area between chainage K1 and chainage K2, ..., the (n-2)th video surveillance area between chainage K(n-3) and chainage K(n-2), L1 / S video surveillance devices are arranged at equal intervals.

[0048] For the (n-1)th video surveillance area between chainage K(n-2) and chainage K(n-2)+L2, L2 / S video surveillance devices are arranged at equal intervals.

[0049] Step 2: When each video monitoring device captures video images, its shooting angle is directly facing the tunnel surface and remains stationary. Its video field of view covers its corresponding video monitoring area. The length of the video monitoring area is equal to the visible range of the video monitoring device, which is S meters; the width of the video monitoring area is equal to the width of the tunnel surface, which is W meters. This ensures the integrity and continuity of the video monitoring images captured by each video monitoring device, and can completely cover the road surface of the tunnel construction area.

[0050] Step 3: For each video surveillance image captured by a video surveillance device, pre-draw virtual segment markers to determine the location of the video surveillance area corresponding to the video surveillance image;

[0051] Specifically, between any two adjacent foundation piles, multiple virtual pile numbers are generated at intervals of S meters within the visible range, and the name of each virtual pile number is related to the mileage.

[0052] For any video surveillance image captured by a video surveillance device, its starting and ending points along the length direction are marked using the foundation pile numbers and / or virtual pile numbers at both ends of the corresponding video surveillance area. The foundation pile numbers and / or virtual pile numbers are superimposed on the starting and ending points of the video surveillance image to form virtual segment marks of the video surveillance image. Through the virtual segment marks, the actual mileage position of the starting and ending points of the video surveillance image within the tunnel can be directly converted.

[0053] For example, for the first video surveillance area between station K0 and station K1, a total of L1 / S virtual station numbers are generated in sequence, namely: station K0+S, station K0+2S, ..., station K0+(L1 / S)*S=K0+L1;

[0054] Therefore, among the L1 / S video surveillance devices arranged in the first video surveillance area, the starting point and ending point markers of the first video surveillance device are respectively: station number K0, station number K0+S; the starting point and ending point markers of the second video surveillance device are respectively: station number K0+S, station number K0+2S; and so on, the starting point and ending point markers of the L1 / S video surveillance devices are respectively: station number K0+L1, station number K1.

[0055] Step 4: Perform video virtual meshing processing on the video surveillance images captured by each video surveillance device to form multiple virtual meshes, and establish the mapping between the position of the virtual mesh in the video surveillance image and the actual position in the physical space of the tunnel.

[0056] For video surveillance images with overlaid virtual segment markers, the images are gridded according to the set virtual grid size to form a virtual grid of a rows and b columns, where a is the number of rows and b is the number of columns. For the virtual grid in the i-th row and j-th column of the video surveillance image, where i∈[1,a] and j∈[1,b], the following algorithm is used to obtain its actual position (x0,y0) in the tunnel physical space, where x0 and y0 are the horizontal and vertical coordinates of its actual position in the tunnel physical space, respectively.

[0057] In the video surveillance image, a rectangular coordinate system is established with the length direction (the longitudinal direction of the tunnel) as the y-axis, the width direction (the transverse direction of the tunnel) as the x-axis, and the lower left corner of the video surveillance image as the origin.

[0058] The virtual segment marker corresponding to the origin of the coordinate system is the foundation pile number and / or virtual pile number marker of the starting point of the video surveillance image. Using this virtual segment marker, the actual mileage position of the origin of the coordinate system within the tunnel is obtained, denoted as M. The actual position (x0, y0) is obtained using the following formula:

[0059] x0=j*g

[0060] y0=M+i*g

[0061] Where: g is the side length of the virtual grid;

[0062] Step 5: When personnel location is required, obtain the video surveillance images captured by each video surveillance device; then, use a target recognition algorithm to identify personnel in each video surveillance image, assuming that a person is identified in video surveillance image P.

[0063] Then, the virtual grid where the person is located is first determined in the video surveillance image P, and the position of the virtual grid is mapped to the actual position in the physical space of the tunnel, thereby achieving accurate positioning of the person.

[0064] When tracking the location and trajectory of personnel, the specific steps include:

[0065] The video surveillance images captured by each video surveillance device at the current moment are used to form a video surveillance image sequence; a target recognition algorithm is used to detect the location of personnel.

[0066] Then, using a target tracking algorithm, the position of the same person in various video surveillance devices at the next moment is continuously tracked to achieve personnel trajectory tracking.

[0067] The following is a specific example:

[0068] Step 1: Install foundation piles and video surveillance equipment.

[0069] The video surveillance devices, which serve as video monitoring points, are fixed bullet cameras and are deployed at equal intervals along the sidewalls of the tunnel's longitudinal axis. Each video surveillance device faces the ground, and the video field of view covers the road inside the tunnel.

[0070] If the tunnel construction progress is X meters, the starting chainage of the construction area is K0, the ending chainage of the construction area is K(X / 1000)+(X%1000), the visible range of each video monitoring device is S meters, and a total of N=X / S video monitoring devices are deployed.

[0071] For example, if the tunnel construction progress is 2300 meters, and the foundation pile spacing is L1 meters (equivalent to 1000 meters), then a total of 4 foundation piles are needed. These foundation piles are numbered as follows: pile number K0, pile number K1, pile number K2, and pile number K2+300 meters; where 300 meters is the value of L2. Pile number K2+300 meters means that the distance between this foundation pile and its adjacent foundation pile at pile number K2 is 300 meters.

[0072] Therefore, the tunnel construction area is divided into three sections: K0~K1, K1~K2, and K2~K2+300 meters.

[0073] The visible range of each video surveillance device is S = 100 meters. A total of 23 video surveillance devices are deployed, with each device placed at equal intervals. Therefore, 10 devices are deployed in the K0-K1 section, 10 devices are deployed in the K1-K2 section, and 3 devices are deployed in the K2-K2+300 meter section.

[0074] The purpose of this step is to ensure that the video monitoring area of ​​each video monitoring device in the tunnel is complete and continuous, covering the entire construction area of ​​the tunnel, and that the use of fixed bullet cameras can maintain the stability of the monitoring image, providing a stable image for subsequent segment marking and virtual grid.

[0075] Step 2: Set up virtual segment markers and virtual areas for each video surveillance device.

[0076] Taking segment K0 to K1 as an example, the visible range of each video surveillance device is S meters. Virtual station numbers K0 and K0+S are superimposed on the video image area of ​​video 1#; virtual station numbers K0+S and K0+2S are superimposed on the video image area of ​​video 2#; this operation is repeated, and the video image area of ​​video N# is K0+(N-1)S and K0+NS. The virtual station numbers drawn and superimposed on the video 1# to video N# are stored to form virtual segment markers along the video area of ​​the tunnel, which can be easily retrieved later.

[0077] After completing the virtual paragraph marking, the virtual area is set. Let the width of the video monitoring image of a single video monitoring device be W and the length be S meters. Therefore, the area of ​​the video monitoring area of ​​each video monitoring device is W*S.

[0078] Step 3: Set up the video virtual grid to create a mapping between the virtual grid and the tunnel's physical space. Let the precision of the virtual grid be g, typically g = 1, meaning 1 square meter is one cell, and the side length of each cell is 1 meter.

[0079] Taking segment K0 to K1 as an example, and video 1# as an example, such as Figure 2 As shown, the actual position (x0, y0) corresponding to each virtual grid P(j, i) is:

[0080] x0=j*g

[0081] y0=M+i*g

[0082] Step 4: Using target recognition algorithms, such as the YOlOv4 algorithm, the identification and tracking of personnel in the tunnel are realized. Then, the center point of the personnel is mapped on a virtual grid, and finally the personnel are located and identified in the tunnel.

[0083] The personnel location and trajectory tracking process includes video sequence analysis, target detection, target tracking, and target feature extraction and recognition steps. For the tunnel video sequence to be detected, the target detection algorithm is first used to detect the personnel in the video; then, the target tracking algorithm is used to continuously track the position of the same person in the video sequence. The output of this position is the center point of the detected personnel target, Person-center(x,y), which is compared with a virtual grid to form a location recognition.

[0084] The present invention proposes a method for locating personnel in tunnels based on video recognition, the main steps of which can be summarized as follows:

[0085] Step 1: Install fixed bullet camera video surveillance devices at equal intervals so that the cameras do not rotate and the coverage area of ​​each video surveillance device is continuous.

[0086] Step 2 involves setting virtual segment markers and virtual areas for each video monitoring device to ensure that the location of the video monitoring screen corresponds one-to-one with the actual tunnel station mileage.

[0087] Step 3: Set up the video virtual grid to map the virtual grid to the physical space of the tunnel, so that the area and width of the video surveillance screen are consistent with the actual tunnel road.

[0088] Step 4: Use target recognition algorithms to identify and track personnel in the tunnel, and then map the center point of the personnel onto a virtual grid to achieve the location and identification of personnel in the tunnel.

[0089] The tunnel personnel positioning method based on video recognition provided by this invention has the following advantages:

[0090] This invention proposes a tunnel personnel positioning method based on video recognition. It eliminates the need for personnel to wear positioning devices and can accurately and quickly determine the location and trajectory of personnel by intuitively and quickly using video monitoring image information. It also reduces the interference of various complex environments in the tunnel on the positioning equipment and has the advantages of high positioning efficiency and accurate positioning results.

[0091] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for locating personnel in tunnels based on video recognition, characterized in that, Includes the following steps: Step 1: Install foundation piles and video surveillance equipment. Along the longitudinal direction of the tunnel, on one side of the tunnel construction area, multiple foundation piles are arranged at intervals, each foundation pile having a foundation pile number related to the mileage; on the other side of the tunnel construction area, multiple video monitoring devices are arranged at intervals. The spacing between the video surveillance devices is equal to the visible range of the video surveillance devices along the longitudinal direction of the tunnel, which is S meters. Step 2: When each video monitoring device captures video images, its shooting angle is directly facing the tunnel surface and remains stationary. Its video field of view covers its corresponding video monitoring area. The length of the video monitoring area is equal to the visible range of the video monitoring device, which is S meters; the width of the video monitoring area is equal to the width of the tunnel surface, which is W meters. This ensures the integrity and continuity of the video monitoring images captured by each video monitoring device, and can completely cover the road surface of the tunnel construction area. Step 3: For each video surveillance image captured by a video surveillance device, pre-draw virtual segment markers to determine the location of the video surveillance area corresponding to the video surveillance image; Specifically, between any two adjacent foundation piles, multiple virtual pile numbers are generated at intervals of S meters within the visible range, and the name of each virtual pile number is related to the mileage. For any video surveillance image captured by a video surveillance device, its starting and ending points along the length direction are marked using the foundation pile numbers and / or virtual pile numbers at both ends of the corresponding video surveillance area. The foundation pile numbers and / or virtual pile numbers are superimposed on the starting and ending points of the video surveillance image to form virtual segment marks of the video surveillance image. Through the virtual segment marks, the actual mileage position of the starting and ending points of the video surveillance image within the tunnel can be directly converted. Step 4: Perform video virtual meshing processing on the video surveillance images captured by each video surveillance device to form multiple virtual meshes, and establish the mapping between the position of the virtual mesh in the video surveillance image and the actual position in the physical space of the tunnel. For video surveillance images with overlaid virtual segment markers, the images are gridded according to the set virtual grid size to form a virtual grid of a rows and b columns, where a is the number of rows and b is the number of columns. For the virtual grid in the i-th row and j-th column of the video surveillance image, where i∈[1,a] and j∈[1,b], the following algorithm is used to obtain its actual position (x0,y0) in the tunnel physical space, where x0 and y0 are the horizontal and vertical coordinates of its actual position in the tunnel physical space, respectively. In the video surveillance image, a rectangular coordinate system is established with the length direction (the longitudinal direction of the tunnel) as the y-axis, the width direction (the transverse direction of the tunnel) as the x-axis, and the lower left corner of the video surveillance image as the origin. The virtual segment marker corresponding to the origin of the coordinate system is the foundation pile number and / or virtual pile number marker of the starting point of the video surveillance image. Using this virtual segment marker, the actual mileage position of the origin of the coordinate system within the tunnel is obtained, denoted as M. The actual position (x0, y0) is obtained using the following formula: x0=j*g y0=M+i*g Where: g is the side length of the virtual grid; Step 5: When personnel location is required, obtain the video surveillance images captured by each video surveillance device; then, use a target recognition algorithm to identify personnel in each video surveillance image, assuming that a person is identified in video surveillance image P. Then, the virtual grid where the person is located is first determined in the video surveillance image P, and the position of the virtual grid is mapped to the actual position in the physical space of the tunnel, thereby achieving accurate positioning of the person.

2. The tunnel personnel positioning method based on video recognition according to claim 1, characterized in that, In step 1, each foundation pile has a foundation pile number related to its mileage, specifically: Along the longitudinal direction of the tunnel, between the starting and ending points of the tunnel construction area, on one side of the tunnel, a total of n foundation piles are arranged, and each foundation pile is numbered as follows: pile number K0, pile number K1, ..., pile number K(n-2), pile number K(n-2)+L2; among them, the foundation piles corresponding to pile number K0, pile number K1, ..., pile number K(n-2) are arranged at equal intervals with an interval of L1 meters; the foundation pile of pile number K(n-2)+L2 represents the distance from the foundation pile of the adjacent pile of pile number K(n-2) in front of it to be L2 meters, where L2≤L1. 3.The tunnel personnel positioning method based on video recognition of claim 2, wherein, L1 meter and L2 meter are both integer multiples of the visible range S meter of the video surveillance device.

4. The tunnel personnel positioning method based on video recognition according to claim 2, characterized in that, For the first video surveillance area between chainage K0 and chainage K1, the second video surveillance area between chainage K1 and chainage K2, ..., the (n-2)th video surveillance area between chainage K(n-3) and chainage K(n-2), L1 / S video surveillance devices are arranged at equal intervals. For the (n-1)th video surveillance area between chainage K(n-2) and chainage K(n-2)+L2, L2 / S video surveillance devices are arranged at equal intervals.

5. The tunnel personnel positioning method based on video recognition according to claim 4, characterized in that, For the first video surveillance area between station K0 and station K1, a total of L1 / S virtual station numbers are generated in sequence, namely: station K0+S, station K0+2S, ..., station K0+(L1 / S)*S=K0+L1; Therefore, among the L1 / S video surveillance devices deployed in the first video surveillance area, the starting point and ending point markers of the first video surveillance device are respectively: station number K0, station number K0+S; the starting point and ending point markers of the second video surveillance device are respectively: station number K0+S, station number K0+2S; and so on, the starting point and ending point markers of the L1 / S video surveillance devices are respectively: station number K0+L1, station number K1.

6. The tunnel personnel positioning method based on video recognition according to claim 1, characterized in that, When tracking the location and trajectory of personnel, the specific steps include: The video surveillance images captured by each video surveillance device at the current moment are used to form a video surveillance image sequence; a target recognition algorithm is used to detect the location of personnel. Then, using a target tracking algorithm, the position of the same person in various video surveillance devices at the next moment is continuously tracked to achieve personnel trajectory tracking.

Images