Substation bird detection and tracking method, device and storage medium based on deep learning

Abstract

The present invention discloses a method, device and storage medium for detecting and tracking birds in a substation based on deep learning. The method comprises: obtaining real-time monitoring video images near the substation; using a deep learning method to identify birds in each video image; performing multi-target tracking on all bird entities identified in the video image; performing single-target tracking on each bird in the video image; calculating the safe distance of each bird according to the coordinate interval of each bird on the most recent frame of the video image and the coordinate interval of the nearest substation equipment on the most recent frame of the video image; if the future flight distance of the bird on the image video exceeds the safe distance of the bird, the current bird is determined to be an expulsion target bird, and the video image with the expulsion target bird marked is sent to an alarm device, and an expulsion instruction is sent to an expulsion device; when the portion of the area occupied before and after the expulsion is reduced exceeds a threshold, the expulsion is determined to be successful. The present invention has high tracking accuracy and tracking speed.

Description

Technical Field

[0001] The present invention relates to bird tracking technology, and in particular to a substation bird detection and tracking method, device and storage medium based on deep learning. Background Art

[0002] The background of bird repelling on power grids mainly stems from the problem of bird interference faced by power systems. As the coverage of power grids expands, it is becoming more and more common for birds to nest, roost and move around on power facilities. However, this behavior can easily cause birds to touch wires, causing short circuits, tripping and other accidents, which not only endangers the lives of birds, but also seriously affects the operational stability of the power grid. In addition, the accumulation of bird droppings and nesting materials can also damage equipment. In order to solve this problem, the power sector has gradually adopted various bird repelling measures, such as installing bird repelling devices and using sound and light bird repelling technology, to ensure the safe operation of the power grid and reduce power accidents caused by bird damage.

[0003] The bird-repelling equipment currently available in the industry is indiscriminate, short-term, and non-targeted. It cannot continuously track and strike birds, so its ability to disperse birds is weak and cannot effectively maintain the safety of the power grid for a long time. Tracking moving targets has always been a difficult and hot issue in the field of machine vision research. In the process of bird identification and tracking, there are many challenges and difficulties, such as false detection of birds, deformation of birds, and excessive flying speed of birds. The target detection and tracking algorithm based on deep learning can avoid these problems to a certain extent, effectively improving efficiency and accuracy.

[0004] However, the inference of the twin network based on deep learning takes a long time and has poor real-time performance. The target box of the kernel correlation filter algorithm is already set during the tracking process, and it does not have the ability to adaptively track the target size. This will cause the target box to drift during the tracking process, resulting in tracking failure and low tracking accuracy. The tracking algorithm based on the Neanshift algorithm does not have a good solution for fast-moving, target occlusion, and target scenes with similar backgrounds. Therefore, the existing algorithms cannot balance tracking accuracy and tracking speed. Summary of the invention

[0005] In view of the problems existing in the prior art, the object of the present invention is to provide a substation bird detection and tracking method, device and storage medium based on deep learning with relatively high tracking accuracy and tracking speed.

[0006] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0007] A substation bird detection and tracking method based on deep learning includes the following steps:

[0008] (1) Obtain real-time surveillance video near the substation and extract several frames of video images arranged in chronological order from the surveillance video;

[0009] (2) Use deep learning methods to identify birds in each video image and identify all bird entities in the video image;

[0010] (3) Perform multi-target tracking on all bird entities identified in the video images, that is, calculate the coordinate interval of each bird entity in each video image;

[0011] (4) Single target tracking is performed for each bird in the video image. That is, the template matching method is used to determine whether the bird entity identified in the video images at different times is the same bird, and the future flight distance of the bird in the video image is estimated based on the coordinate interval of the same bird in the video images at different times.

[0012] (5) Calculate the safety distance of each bird based on the coordinate interval of each bird in the most recent video image and the coordinate interval of the nearest substation equipment in the most recent video image, where the safety distance is positively correlated with the coordinate interval of the bird in the video image and the coordinate interval of the nearest substation equipment in the video image;

[0013] (6) If the future flight distance of the bird in the image video exceeds the safe distance of the bird, the current bird is determined to be an expulsion target bird, and the video image with the expulsion target bird marked is sent to the alarm device, and an expulsion instruction is sent to the expulsion device, so that the expulsion device issues an expulsion action to the expulsion target bird;

[0014] (7) Obtaining the area occupied by the expelled target bird in the video image after the preset time period of the expulsion instruction. If the reduction in the occupied area before and after the expulsion exceeds a threshold, the expulsion is determined to be successful.

[0015] Furthermore, step (1) specifically includes:

[0016] Obtain real-time surveillance video captured by surveillance cameras near the substation;

[0017] The video frames of the monitoring video are sampled at equal intervals, and the sampled video frames are arranged in chronological order as video images arranged in chronological order.

[0018] Furthermore, step (2) specifically includes:

[0019] According to the mapping rule between the three-dimensional real space and the two-dimensional image space, a bird monitoring image area corresponding to the preset three-dimensional bird monitoring space range in the video image is obtained;

[0020] A bird recognition model based on a deep neural network is used to recognize all bird entities in the video image and generate candidate bird recognition frames, wherein the bird recognition model is trained by multiple images containing birds and labeled images with bird recognition frames marked;

[0021] The candidate recognition frames that are not located in the bird monitoring image area are deleted, and the remaining candidate recognition frames are used as the recognition frames of the final recognized bird entities.

[0022] Furthermore, step (3) specifically includes:

[0023] Establish a two-dimensional coordinate system including x and y directions on the video image;

[0024] Generate a number of equally spaced straight lines in the x and y directions of the two-dimensional coordinate system of each video image, and use the intersection of the straight lines in the x and y directions as the positioning point;

[0025] Fitting the outer contour of the bird in each bird entity recognition frame in each video image to obtain the outer contour of each bird entity;

[0026] All the positioning points covered by the outer contour of each bird entity are obtained, and the coordinate set of all the positioning points is used as the coordinate interval of the current bird entity on the current video image.

[0027] Furthermore, step (4) specifically includes:

[0028] Obtaining the identification box of each bird entity in the video image identified by the deep learning method;

[0029] Allocate unique identification numbers to all bird entities appearing in the first frame of the video image, and store the image within the identification frame of each bird entity as a template corresponding to the unique identification number;

[0030] Match the image in the recognition frame of the bird entity identified in the video image of the subsequent frame with all stored templates, and mark the unique identification number of the bird in the recognition frame that successfully matches the template as the unique identification number consistent with the template; if no match is successful, store the image in the recognition frame of the bird entity as a new template;

[0031] The bird entities with the same unique identification number in the video images of different frames are regarded as the same bird. According to the coordinate range of the same bird in the video images at different times, the future flight distance of the bird in the image video is estimated.

[0032] Furthermore, the method of estimating the future flight distance of the bird on the video image according to the coordinate intervals of the same bird on the video image at different times specifically includes:

[0033] Get the coordinate interval of the same bird in the video images of the three nearest adjacent frames;

[0034] According to the coordinate interval of the bird on the three adjacent frames of video images, the first image distance d1 moved by the bird when it moves from the first frame of video image to the second frame of video image in the three adjacent frames of video images, and the second image distance d2 moved by the bird when it moves from the second frame of video image to the third frame of video image are calculated:

[0035] d1=p1s,

[0036] d2=p2s,

[0037] Where p1 represents the number of positioning points that the bird moves from the first frame of video image to the second frame of video image, p2 represents the number of positioning points that the bird moves from the second frame of video image to the third frame of video image, and s represents the distance between adjacent positioning points;

[0038] Obtain the coordinate interval of the nearest substation equipment in three adjacent frames of video images, and calculate the first image distance l1 of the substation equipment when it moves from the first frame of video image to the second frame of video image in the three adjacent frames of video images, and the second image distance l2 of the substation equipment when it moves from the second frame of video image to the third frame of video image:

[0039] l1=q1s,

[0040] l2=q2s,

[0041] Where, q1 represents the number of positioning points moved when the nearest substation equipment moves from the first frame of video image to the second frame of video image, and q2 represents the number of positioning points moved when the nearest substation equipment moves from the second frame of video image to the third frame of video image;

[0042] The future flight distance of the bird on the image video is calculated according to the following formula:

[0043] ,

[0044] In the formula, B represents the future flight distance of the bird in the image video, β represents the conversion coefficient, and t represents the time interval between adjacent frames of the video image. When the bird is accelerating, the value of is negative, and when it is decelerating, the value is positive.

[0045] Furthermore, step (5) specifically includes:

[0046] According to the coordinate interval of the current bird in the most recent frame of the video and the coordinate interval of the nearest substation equipment in the most recent frame of the video, the safety distance of the current bird is calculated according to the following formula:

[0047] L = αN1N2,

[0048] N1=n1s,

[0049] N2=n2s,

[0050] Where L is the current bird's safety distance, α is the conversion coefficient between safety distance and shape size, N1 is the shape size of the current bird in the latest frame of image video, N2 is the shape size of the nearest substation equipment in the latest frame of image video, s is the distance between adjacent positioning points, n1 is the number of positioning points in the coordinate interval of the current bird in the latest frame of image video, and n2 is the number of positioning points in the coordinate interval of the nearest substation equipment in the latest frame of image video.

[0051] Furthermore, the coordinate interval of the nearest substation equipment on the video image is obtained by the following method:

[0052] A substation equipment recognition model based on a deep neural network is used to recognize all substation equipment in the video image and generate a substation equipment recognition frame, wherein the substation equipment recognition model is trained by multiple images containing substation equipment and labeled images with substation equipment recognition frames marked;

[0053] Fitting the outer contour of each substation equipment in each substation equipment identification frame in the video image to obtain the outer contour of each substation equipment;

[0054] Calculate the distance between the outer contour of all substation equipment and the outer contour of the current bird, and select the substation equipment with the shortest distance as the nearest substation equipment;

[0055] All positioning points covered by the outer contour of the nearest substation equipment are obtained, and the coordinate set of all positioning points is used as the coordinate interval of the nearest substation equipment on the video image.

[0056] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above method.

[0057] A computer-readable storage medium having a computer program / instruction stored thereon, wherein the computer program / instruction implements the above method when executed by a processor.

[0058] Compared with the prior art, the present invention has the following beneficial effects: the present invention adopts a deep learning method to identify birds, then simultaneously performs multi-target tracking on multiple birds in the same video image, and then performs single-target tracking on a single bird, with high tracking accuracy and low target loss. The present invention adopts real-time video processing, with high real-time performance and fast tracking speed, and can detect dangerous birds more quickly and expel them. The present invention can be applied to target identification and tracking in various scenarios, including but not limited to power grids, farmland bird repelling, high-altitude drone positioning, and military target identification and tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of a method for detecting and tracking birds in a substation based on deep learning provided by an embodiment of the present invention;

[0060] Figure 2 It is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0062] The following is Example 1. The embodiment of the present invention provides a method for detecting and tracking birds in a substation based on deep learning. Figure 1 As shown, the following steps are included:

[0063] S101. Acquire a real-time monitoring video near a substation, and extract a plurality of frames of video images arranged in chronological order from the monitoring video.

[0064] In the specific implementation, the real-time monitoring video taken by the monitoring near the substation is first obtained, for example, the monitoring video of the most recent minute; then the video frames of the monitoring video are sampled at equal intervals, and the sampled video frames are arranged in chronological order as video images arranged in chronological order.

[0065] S102: Use a deep learning method to identify birds in each video image, and identify all bird entities in the video image.

[0066] This step specifically includes:

[0067] According to the mapping rule between the three-dimensional real space and the two-dimensional image space, a bird monitoring image area corresponding to the preset three-dimensional bird monitoring space range in the video image is obtained; wherein the mapping rule between the three-dimensional real space and the two-dimensional image space is an existing technology and will not be described in detail;

[0068] A bird recognition model based on a deep neural network is used to recognize all bird entities in the video image and generate candidate bird recognition frames, wherein the bird recognition model is trained by multiple images containing birds and labeled images with bird recognition frames marked; the bird recognition model can be various types of deep neural networks, such as a MicroNet lightweight network;

[0069] The candidate recognition frames that are not in the bird monitoring image area are deleted, and the remaining candidate recognition frames are used as the recognition frames of the bird entities that are finally recognized. Birds that are not in the bird monitoring image area are not tracked, which can reduce the tracking difficulty and reduce the algorithm space.

[0070] S103: Perform multi-target tracking on all bird entities identified in the video image.

[0071] Multi-target tracking is to calculate the coordinate interval of each bird entity in each video image. It includes the following steps:

[0072] Establish a two-dimensional coordinate system including x and y directions on the video image;

[0073] Generate a number of equally spaced straight lines in the x and y directions of the two-dimensional coordinate system of each video image, and use the intersection of the straight lines in the x and y directions as the positioning point;

[0074] Fitting the outer contour of the bird in each bird entity recognition frame in each video image to obtain the outer contour of each bird entity; the fitting may be performed by using a small square approximation method in the recognition frame to fit the outer contour, or other methods may be used;

[0075] All the positioning points covered by the outer contour of each bird entity are obtained, and the coordinate set of all the positioning points is used as the coordinate interval of the current bird entity on the current video image.

[0076] S104: Perform single target tracking on each bird in the video image.

[0077] Single target tracking is to determine whether the bird entity identified in the video images at different times is the same bird based on the template matching method, and estimate the bird's future flight distance in the video image based on the coordinate range of the same bird in the video images at different times.

[0078] The specific steps of single target basis include:

[0079] Obtaining the identification box of each bird entity in the video image identified by the deep learning method;

[0080] Allocate unique identification numbers to all bird entities appearing in the first frame of the video image, and store the image within the identification frame of each bird entity as a template corresponding to the unique identification number;

[0081] Match the image in the recognition frame of the bird entity identified in the video image of the subsequent frame with all stored templates, and mark the unique identification number of the bird in the recognition frame that successfully matches the template as the unique identification number consistent with the template; if no match is successful, store the image in the recognition frame of the bird entity as a new template;

[0082] The bird entities with the same unique identification number in the video images of different frames are regarded as the same bird. According to the coordinate range of the same bird in the video images at different times, the future flight distance of the bird in the image video is estimated.

[0083] Among them, the specific method for estimating the future flight distance of the bird on the image video is:

[0084] Get the coordinate interval of the same bird in the video images of the three nearest adjacent frames;

[0085] According to the coordinate intervals of the bird on three adjacent frames of video images, the first image distance d1 moved by the bird from the first frame of video image to the second frame of video image, and the second image distance d2 moved from the second frame of video image to the third frame of video image are calculated:

[0086] d1=p1s,

[0087] d2=p2s,

[0088] Where p1 represents the number of positioning points that the bird moves from the first frame of video image to the second frame of video image, p2 represents the number of positioning points that the bird moves from the second frame of video image to the third frame of video image, and s represents the distance between adjacent positioning points;

[0089] Get the coordinate interval of the nearest substation equipment in three adjacent frames of video images, and calculate the first image distance l1 when the substation equipment moves from the first frame of video image to the second frame of video image, and the second image distance l2 when the substation equipment moves from the second frame of video image to the third frame of video image:

[0090] l1=q1s,

[0091] l2=q2s,

[0092] Where, q1 represents the number of positioning points moved when the nearest substation equipment moves from the first frame of video image to the second frame of video image, and q2 represents the number of positioning points moved when the nearest substation equipment moves from the second frame of video image to the third frame of video image;

[0093] The future flight distance of the bird on the image video is calculated according to the following formula:

[0094] ,

[0095] In the formula, B represents the future flight distance of the bird in the image video, β represents the conversion coefficient, and t represents the time interval between adjacent frames of the video image. When the bird is accelerating, the value of is negative, and when it is decelerating, the value is positive.

[0096] S105, calculating the safety distance of each bird according to the coordinate interval of each bird in the latest frame of video image and the coordinate interval of the nearest substation equipment in the latest frame of video image.

[0097] The safety distance is positively correlated with the coordinate interval of the bird on the video image and the coordinate interval of the nearest substation equipment on the video image. The safety distance calculation method specifically includes: according to the coordinate interval of the current bird on the most recent frame of the video image and the coordinate interval of the nearest substation equipment on the most recent frame of the video image, the safety distance of the current bird is calculated according to the following formula:

[0098] L = αN1N2,

[0099] N1=n1s,

[0100] N2=n2s,

[0101] Where L is the current bird's safety distance, α is the conversion coefficient between safety distance and shape size, N1 is the shape size of the current bird in the latest frame of image video, N2 is the shape size of the nearest substation equipment in the latest frame of image video, s is the distance between adjacent positioning points, n1 is the number of positioning points in the coordinate interval of the current bird in the latest frame of image video, and n2 is the number of positioning points in the coordinate interval of the nearest substation equipment in the latest frame of image video.

[0102] Among them, the coordinate interval of the nearest substation equipment on the video image is obtained by the following method:

[0103] A substation equipment recognition model based on a deep neural network is used to recognize all substation equipment in the video image and generate a substation equipment recognition frame, wherein the substation equipment recognition model is trained by multiple images containing substation equipment and labeled images with substation equipment recognition frames marked; the substation equipment recognition model may be consistent with the bird recognition model structure or inconsistent, and may be any other deep neural network;

[0104] Fitting the outer contour of each substation equipment in each substation equipment identification frame in the video image to obtain the outer contour of each substation equipment;

[0105] Calculate the distance between the outer contour of all substation equipment and the outer contour of the current bird, and select the substation equipment with the shortest distance as the nearest substation equipment;

[0106] All positioning points covered by the outer contour of the nearest substation equipment are obtained, and the coordinate set of all positioning points is used as the coordinate interval of the nearest substation equipment on the video image.

[0107] S106. If the future flight distance of the bird in the image video exceeds the safe distance of the bird, the current bird is determined to be an expulsion target bird, a video image with the expulsion target bird marked is sent to an alarm device, and an expulsion instruction is sent to an expulsion device, so that the expulsion device issues an expulsion action to the expulsion target bird.

[0108] Specifically, the expulsion device may be various types of expulsion devices, such as laser expulsion devices, sound expulsion devices, etc., and the alarm device may be central control equipment, etc.

[0109] S107, obtaining the area occupied by the expelled target bird in the video image after the preset time period of sending the expulsion instruction, if the reduction of the occupied area before and after the expulsion exceeds a threshold, it is determined that the expulsion is successful.

[0110] If the bird flies farther and farther, the area will become smaller and smaller. When the reduced part exceeds the preset threshold, it is determined that the expulsion is successful. Otherwise, it is determined that the expulsion is unsuccessful and the expulsion instruction continues to be sent to the expulsion device.

[0111] The following is Example 2. Figure 2 is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention, and the embodiment of the present invention provides services for implementing the method of the above-mentioned embodiment 1 of the present invention. Figure 2 As shown, the device may include: a memory 201 storing a computer executable program; a processor 202 coupled to the memory 201; the processor 202 calls the computer executable program stored in the memory 201 to execute the steps in the method described in the first embodiment.

[0112] The memory 201 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the memory 201 may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as "hard drives"). A program / utility having a set (at least one) of program modules may be stored in, for example, the memory 201, such program modules including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include the implementation of a network environment. The computer executable program of the program module typically performs the functions and / or methods in the embodiments described in the present invention.

[0113] The processor 202 executes various functional applications and data processing by running the programs stored in the memory 201, such as implementing the method provided in the first embodiment of the present invention.

[0114] The code of the computer executable program can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" language or similar programming languages.

[0115] The following is Example 3. The embodiment of the present invention provides a storage medium including a computer executable program, and the computer executable program is used to execute the method of Example 1 when executed by a computer processor.

[0116] The storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device.

[0117] Of course, the storage medium including a computer executable program provided by an embodiment of the present invention, the computer executable program is not limited to the above method operations, and can also execute related operations in the method provided by any embodiment of the present invention.

[0118] It should be understood that the above embodiments and descriptions only describe the principles, main features and advantages of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, and these changes and improvements all fall within the scope of protection of the present invention.

Claims

1. A substation bird detection and tracking method based on deep learning, characterized in that: The steps include: (1) Obtain real-time surveillance video near the substation and extract several frames of video images arranged in chronological order from the surveillance video; (2) Use deep learning methods to identify birds in each video image and identify all bird entities in the video image; (3) Perform multi-target tracking on all bird entities identified in the video images, that is, calculate the coordinate interval of each bird entity in each video image; (4) Single target tracking is performed for each bird in the video image. That is, the template matching method is used to determine whether the bird entity identified in the video images at different times is the same bird, and the future flight distance of the bird in the video image is estimated based on the coordinate interval of the same bird in the video images at different times. (5) Calculate the safety distance of each bird based on the coordinate interval of each bird in the most recent video image and the coordinate interval of the nearest substation equipment in the most recent video image, where the safety distance is positively correlated with the coordinate interval of the bird in the video image and the coordinate interval of the nearest substation equipment in the video image; (6) If the future flight distance of the bird in the image video exceeds the safe distance of the bird, the current bird is determined to be an expulsion target bird, and the video image with the expulsion target bird marked is sent to the alarm device, and an expulsion instruction is sent to the expulsion device, so that the expulsion device issues an expulsion action to the expulsion target bird; (7) Obtaining the area occupied by the expelled target bird in the video image after the preset time period of the expulsion instruction. If the reduction in the occupied area before and after the expulsion exceeds a threshold, the expulsion is determined to be successful; The method of estimating the future flight distance of the bird on the video image according to the coordinate intervals of the same bird on the video image at different times specifically includes: Get the coordinate interval of the same bird in the video images of the three nearest adjacent frames; According to the coordinate interval of the bird on the three adjacent frames of video images, the first image distance d1 moved by the bird when it moves from the first frame of video image to the second frame of video image in the three adjacent frames of video images, and the second image distance d2 moved by the bird when it moves from the second frame of video image to the third frame of video image are calculated: d1=p1s, d2=p2s, Where p1 represents the number of positioning points that the bird moves from the first frame of video image to the second frame of video image, p2 represents the number of positioning points that the bird moves from the second frame of video image to the third frame of video image, and s represents the distance between adjacent positioning points; Obtain the coordinate interval of the nearest substation equipment in three adjacent frames of video images, and calculate the first image distance l1 of the substation equipment when it moves from the first frame of video image to the second frame of video image in the three adjacent frames of video images, and the second image distance l2 of the substation equipment when it moves from the second frame of video image to the third frame of video image: l1=q1s, l2=q2s, Where, q1 represents the number of positioning points moved when the nearest substation equipment moves from the first frame of video image to the second frame of video image, and q2 represents the number of positioning points moved when the nearest substation equipment moves from the second frame of video image to the third frame of video image; The future flight distance of the bird on the image video is calculated according to the following formula: ， In the formula, B represents the future flight distance of the bird in the image video, β represents the conversion coefficient, and t represents the time interval between adjacent frames of the video image. When the bird is accelerating The value is negative, and the deceleration is The value of is positive.

2. The substation bird detection and tracking method based on deep learning according to claim 1 is characterized in that: Step (1) specifically includes: Obtain real-time surveillance video captured by surveillance cameras near the substation; The video frames of the monitoring video are sampled at equal intervals, and the sampled video frames are arranged in chronological order as video images arranged in chronological order.

3. The substation bird detection and tracking method based on deep learning according to claim 1 is characterized in that: Step (2) specifically includes: According to the mapping rule between the three-dimensional real space and the two-dimensional image space, a bird monitoring image area corresponding to the preset three-dimensional bird monitoring space range in the video image is obtained; A bird recognition model based on a deep neural network is used to recognize all bird entities in the video image and generate candidate bird recognition frames, wherein the bird recognition model is trained by multiple images containing birds and labeled images with bird recognition frames marked; The candidate recognition frames that are not located in the bird monitoring image area are deleted, and the remaining candidate recognition frames are used as the recognition frames of the final recognized bird entities.

4. The substation bird detection and tracking method based on deep learning according to claim 1 is characterized in that: Step (3) specifically includes: Establish a two-dimensional coordinate system including x and y directions on the video image; Generate a number of equally spaced straight lines in the x and y directions of the two-dimensional coordinate system of each video image, and use the intersection of the straight lines in the x and y directions as the positioning point; Fitting the outer contour of the bird in each bird entity recognition frame in each video image to obtain the outer contour of each bird entity; All the positioning points covered by the outer contour of each bird entity are obtained, and the coordinate set of all the positioning points is used as the coordinate interval of the current bird entity on the current video image.

5. The substation bird detection and tracking method based on deep learning according to claim 1 is characterized in that: Step (4) specifically includes: Obtaining the identification box of each bird entity in the video image identified by the deep learning method; Allocate unique identification numbers to all bird entities appearing in the first frame of the video image, and store the image within the identification frame of each bird entity as a template corresponding to the unique identification number; Match the image in the recognition frame of the bird entity identified in the video image of the subsequent frame with all stored templates, and mark the unique identification number of the bird in the recognition frame that successfully matches the template as the unique identification number consistent with the template; if no match is successful, store the image in the recognition frame of the bird entity as a new template; The bird entities with the same unique identification number in the video images of different frames are regarded as the same bird. According to the coordinate range of the same bird in the video images at different times, the future flight distance of the bird in the image video is estimated.

6. The substation bird detection and tracking method based on deep learning according to claim 1 is characterized in that: Step (5) specifically includes: According to the coordinate interval of the current bird in the most recent frame of the video and the coordinate interval of the nearest substation equipment in the most recent frame of the video, the safety distance of the current bird is calculated according to the following formula: L = αN1N2, N1=n1s, N2=n2s, Where L is the current bird's safety distance, α is the conversion coefficient between safety distance and shape size, N1 is the shape size of the current bird in the latest frame of image video, N2 is the shape size of the nearest substation equipment in the latest frame of image video, s is the distance between adjacent positioning points, n1 is the number of positioning points in the coordinate interval of the current bird in the latest frame of image video, and n2 is the number of positioning points in the coordinate interval of the nearest substation equipment in the latest frame of image video.

7. The substation bird detection and tracking method based on deep learning according to claim 1 is characterized in that: The coordinate interval of the nearest substation equipment on the video image is obtained by the following method: A substation equipment recognition model based on a deep neural network is used to recognize all substation equipment in the video image and generate a substation equipment recognition frame, wherein the substation equipment recognition model is trained by multiple images containing substation equipment and labeled images with substation equipment recognition frames marked; Fitting the outer contour of each substation equipment in each substation equipment identification frame in the video image to obtain the outer contour of each substation equipment; Calculate the distance between the outer contour of all substation equipment and the outer contour of the current bird, and select the substation equipment with the shortest distance as the nearest substation equipment; All positioning points covered by the outer contour of the nearest substation equipment are obtained, and the coordinate set of all positioning points is used as the coordinate interval of the nearest substation equipment on the video image.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: The processor executes the computer program to implement the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: The computer program / instructions, when executed by a processor, implement the method of any one of claims 1-7.

Images