A multi-target tracking method for intelligent substation field monitoring based on YOLO algorithm

Abstract

This invention proposes a multi-target tracking method for on-site monitoring of intelligent substations based on the YOLO algorithm. It utilizes the YOLO algorithm to detect all moving objects, achieving fast and accurate target detection. For objects with severe occlusion, an AKCF (Enhanced Kernel Correlation Filter) is proposed to improve algorithm performance. This method tracks targets by minimizing the distance between target features and candidate region features, thus solving the problem of severe target occlusion. Furthermore, a two-stage target association method based on priority matching and motion estimation-based re-matching is proposed to increase tracking accuracy. This invention demonstrates high accuracy and robustness in multi-target tracking, capable of handling complex real-world scenarios, and has broad application prospects in safety monitoring and defect detection of intelligent substations.

Description

Technical Field

[0001] This invention relates to the field of intelligent power system technology, and in particular to a multi-target tracking method for on-site monitoring of intelligent substations based on the YOLO algorithm. Background Technology

[0002] With the development of smart grids, substations, as a crucial component of the power grid, have become paramount in terms of safe operation and monitoring. However, current intelligent monitoring technologies and security measures for substations face a series of problems and challenges. The most significant issue is the accurate detection and tracking of multiple moving objects within a given scene. Due to the complex environment, diverse scenarios, and challenging lighting conditions of substations, current substation monitoring technologies often suffer from issues such as low accuracy and robustness, susceptibility to environmental interference, and long processing times. The YOLO algorithm is a high-precision and fast target detection algorithm capable of rapidly detecting and locating all targets in a scene. However, due to the complexity of substation monitoring scenarios, simply using the YOLO algorithm cannot solve the problem of target recognition and tracking. Issues such as occlusion, lighting effects, and target motion all contribute to difficulties in target recognition and tracking. Summary of the Invention

[0003] In view of this, this invention addresses the problems of low accuracy and robustness, susceptibility to interference from environmental conditions and other factors, and long processing time in traditional substation monitoring technologies. It proposes a multi-target tracking method for intelligent substation field monitoring based on the YOLO algorithm. The YOLO algorithm is used to detect all moving objects, achieving fast and accurate target detection. For objects with severe occlusion, an AKCF (Enhanced Kernel Correlation Filter) is proposed to improve algorithm performance. This method tracks targets by minimizing the distance between target features and candidate region features, thus solving the problem of severe target occlusion. Furthermore, a two-stage target association method based on priority matching and motion estimation-based re-matching is proposed to increase tracking accuracy.

[0004] This invention first utilizes the YOLO algorithm to detect all moving objects, achieving fast and accurate target detection. For objects with severe occlusion, it proposes using the AKCF (Enhanced Kernel Correlation Filter) to improve algorithm performance. This filter tracks the target by minimizing the distance between the target features and candidate region features, thus solving the problem of severe target occlusion. Furthermore, a two-stage target association method based on priority matching and motion estimation-based re-matching is proposed to increase tracking accuracy. Each target has a priority; low-priority targets are canceled from matching first, thus prioritizing the matching results of high-priority targets. Simultaneously, motion estimation techniques are used to correct the matching results, eliminating drift during tracking. This invention exhibits high accuracy and robustness in multi-target tracking, capable of handling complex real-world scenarios, and has broad application prospects in areas such as safety monitoring and defect detection in smart substations.

[0005] The present invention specifically adopts the following technical solution:

[0006] A multi-target tracking method for on-site monitoring of intelligent substations based on the YOLO algorithm, characterized by the following steps:

[0007] Step S1: Use the YOLO algorithm to perform target detection on the video of the substation site, including dividing the image into multiple grid cells and performing regression on each cell to predict the bounding box and class probability, so as to achieve accurate detection of multiple moving targets in the substation site;

[0008] Step S2: Target tracking is performed using the enhanced kernel correlation filter AKCF, where AKCF is a sample-trained filter that achieves accurate target tracking by minimizing the distance between target features and candidate region features.

[0009] Step S3: A two-stage target association method based on priority matching and motion estimation rematch is adopted to increase tracking accuracy;

[0010] Step S4: Use priority matching to filter and sort the tracking results. Sort the tracking results according to the priority of the target and retain the tracking results of high priority targets.

[0011] Step S5: Correct and match the target using a motion estimation-based rematching method. Correct and match the target's position based on its motion trajectory information to reduce tracking error and improve tracking accuracy.

[0012] The priority matching and motion estimation-based re-matching methods described above further improve the effectiveness and robustness of target tracking, making the method applicable to the accurate tracking of multiple targets in the field monitoring of smart substations. This provides an effective technical means to improve the safety, operation and maintenance efficiency, and reduce the cost of manual intervention in substations.

[0013] Further, in step S4, it is assumed that in frame t, the M detection results are represented as the set of detected targets DET = {p1, p2, ..., p...} M}, and the tracker's prediction results for N targets are represented as the tracked target set TRA={q1,q2,…,q M}; Define UMT as a set of mismatched tracking targets, MP as a set of matching pairs; pv represents the maximum response value predicted by AKCF; IOU represents the degree of overlap between the tracking target frame and the detection target frame, calculated as follows:

[0014]

[0015] Where S t S represents the target frame being tracked. d Indicates the detection target frame

[0016] The two-dimensional overlap correlation matrix IM is used to represent the similarity between targets:

[0017]

[0018] Among them, IOU mn Represents the nth tracking target q in the tracking set TRA n With the m-th detection target p in the detection target set DET m The overlap of target frames between them.

[0019] Further, in step S5, the goal of object association is to achieve object re-matching after long-term occlusion; the state of the target is represented by the position, average velocity, and size of the target frame; tracking target q i The state in frame t is represented as follows:

[0020]

[0021] where i∈N t N t It is the number of filters for the mismatched tracking target set UMT in frame t. Representing the target q i Location coordinates, Indicates the target size. This represents the average velocity of the three frames before the occlusion.

[0022] During the target's movement, the geometric transformation characteristics of the target are used to predict the target's position and trajectory, thereby eliminating the drift phenomenon during the tracking process. By collecting the target's motion information and combining it with a weighted compensation method for re-matching, that is, the target's position is corrected and matched according to the target's motion trajectory information, so as to minimize the tracking error.

[0023] In this invention, priority matching and motion estimation-based rematching methods are combined to achieve accurate tracking of multiple targets at the substation site, improving safety monitoring effectiveness and operation and maintenance efficiency; and the weights and matching strategies of the priority matching method can be adjusted and modified according to actual needs.

[0024] Multi-target tracking can be applied to the safety monitoring of power equipment and facilities in smart substations, such as the safety monitoring and fault detection of high-voltage switches, transformers, power equipment and distribution lines in substations;

[0025] By employing the YOLO algorithm for target detection and AKCF for target tracking, combined with a two-stage target association method of priority matching and motion estimation-based re-matching, the accuracy and robustness of target tracking can be improved. This invention can enhance the intelligence level of substation monitoring equipment, improve monitoring accuracy and real-time performance, help ensure the safe and stable operation of the power grid, and has higher robustness. This technology has good practicality, foresight, and innovation, and can contribute to the safe and stable operation of smart substations and the power grid.

[0026] Furthermore, a multi-target tracking system for on-site monitoring of intelligent substations based on the YOLO algorithm is characterized by comprising the following modules for executing the aforementioned multi-target tracking method for on-site monitoring of intelligent substations based on the YOLO algorithm:

[0027] Grid cells used to divide video images from substation sites;

[0028] Module used for regressing target bounding boxes and class probabilities;

[0029] This module is used to calculate the distance between target features and candidate region features;

[0030] This module is used to select high-priority targets for tracking.

[0031] A module used for correcting target position and matching.

[0032] In the solution provided by this invention, the YOLO algorithm is first used to detect targets in the video of the substation site during the target detection stage. The YOLO algorithm is characterized by high accuracy and speed, and can quickly and accurately detect all moving objects in the substation site under scenarios with high real-time requirements. By processing and analyzing the video frames, information such as the position, size, and category of the target objects is detected.

[0033] Furthermore, to address the problem of severe target occlusion, this invention employs an Enhanced Kernel Correlation Filter (AKCF) for target tracking. AKCF is a tracker based on a kernel correlation filter that minimizes the distance between target features and candidate region features, thereby achieving accurate target tracking. By comparing and matching the target's appearance information with the features of its neighboring regions, accurate target tracking is possible even under severe occlusion.

[0034] Furthermore, a two-stage target association method is proposed to further improve the accuracy of target tracking. First, the tracking results are sorted according to the priority of the targets, with high-priority targets being retained first to ensure accurate tracking of key targets. When multiple targets appear simultaneously, the tracking results are filtered and sorted through priority matching. This ensures that high-priority targets are matched first, improving tracking accuracy.

[0035] Furthermore, building upon the first stage, a re-matching method based on motion estimation is used to further correct and match the target. During the target's movement, its geometric transformation features can be used to predict its position and trajectory, thereby eliminating drift during tracking. By collecting the target's motion information and combining it with a weighted compensation approach for re-matching—that is, correcting and matching the target's position based on its trajectory information—tracking errors are minimized.

[0036] Compared with the prior art, the present invention and its preferred embodiments have the following beneficial effects:

[0037] 1. Using the YOLO algorithm for target detection and AKCF for target tracking, this method significantly improves both detection speed and tracking accuracy compared to traditional monitoring technologies. It effectively solves the problem of traditional monitoring technologies struggling to detect and track targets in complex environments.

[0038] 2. By using deep learning-based target detection and tracking technology, this invention enables intelligent monitoring equipment. In actual real-time monitoring, this technology helps monitoring equipment to quickly and accurately detect and track targets, thereby reducing the cost of manual intervention and improving monitoring efficiency.

[0039] 3. Improving the accuracy and real-time performance of power equipment monitoring is a crucial means to ensure the safe and stable operation of the power grid. This invention, by enhancing the monitoring accuracy and real-time performance of the intelligent substation monitoring system, can identify abnormal events in real time, thereby improving power grid security and effectively preventing various safety accidents.

[0040] 4. AKCF is employed for target tracking to address the problem of severe target occlusion. A two-stage target association method based on priority matching and motion estimation-based rematching is proposed. These methods can still track targets simply and reliably even when occlusion occurs, thus enhancing the robustness of the monitoring system.

[0041] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0042] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0043] Figure 1 This is a flowchart of a method according to an embodiment of the present invention.

[0044] Figure 2 This is a schematic diagram of the network structure of YOLOV5 according to the present invention. Detailed Implementation

[0045] In the following, specific embodiments of this application will be described in detail with reference to the accompanying drawings. Based on these detailed descriptions, those skilled in the art will be able to clearly understand and implement this application. Without departing from the principles of this application, features from various embodiments can be combined to obtain new implementations, or certain features from some embodiments can be substituted to obtain other preferred implementations.

[0046] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0047] To make the features and advantages of this patent more apparent and understandable, specific embodiments are provided below for detailed explanation:

[0048] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0049] like Figure 1 As shown, this invention provides a multi-target tracking method for on-site monitoring of intelligent substations based on the YOLO algorithm. By employing the YOLO algorithm for target detection and AKCF for target tracking, combined with a two-stage target association method of priority matching and motion estimation-based re-matching, the accuracy and robustness of target tracking can be improved. This invention can enhance the intelligence level of substation monitoring equipment, improve monitoring accuracy and real-time performance, help ensure the safe and stable operation of the power grid, and has higher robustness. This technology has good practicality, forward-looking nature, and innovation, and can contribute to the safe and stable operation of intelligent substations and the power grid.

[0050] In this embodiment, deep learning-based multi-target tracking comprises three parts: target detection, target tracking, and target association, with target association being the core issue. In environments with noise pollution, target detectors may suffer from detection errors, missed detections, and multiple target detections. The purpose of association is to link uncertain observations with the tracking trajectory.

[0051] First, the YOLOv5 algorithm was used to detect objects in the video footage from the substation site. YOLOv5, as a real-time object detection algorithm, is characterized by high accuracy and speed. It transforms the object detection problem into a regression problem by dividing the image into multiple grid cells, with each cell predicting a series of bounding boxes and corresponding class probabilities, thereby enabling the detection of multiple moving targets in the substation site. The constraints of YOLOv5 limit the number of surrounding targets that can be predicted, especially small objects. Figure 2 This is a schematic diagram of the YOLOv5 network structure.

[0052] Furthermore, an enhanced kernel correlation filter (AKCF) is employed for target tracking. By comparing and matching the target's appearance information with features of its neighboring regions, accurate target tracking is possible even when the target is severely occluded.

[0053] Furthermore, a two-stage target association method is proposed to further improve the accuracy of target tracking. Assume that in frame t, the M detection results are represented as the set of detected targets DET = {p1, p2, ..., p...}. M}, and the tracker's prediction results for N targets are represented as the tracked target set TRA={q1,q2,…,q M}. UMT is defined as a set of mismatched tracked targets, and MP is a set of matched pairs. pv represents the maximum response value predicted by AKCF. IOU represents the degree of overlap between the tracked target frame and the detected target frame, calculated as follows:

[0054]

[0055] Where S t S represents the target frame being tracked. d Indicates the detection target frame

[0056] A two-dimensional overlap correlation matrix IM is used to represent the similarity between targets.

[0057]

[0058] Among them, IOU mn Represents the nth tracking target q in the tracking set TRA n With the m-th detection target p in the detection target set DET m The overlap of target frames between them.

[0059] The goal of the second-stage object association is to achieve re-matching of objects that reappear after long-term occlusion. Considering that the target's motion state in adjacent frames can be approximated as uniform motion, and that the target frame changes are small, the target's state is represented by its position, average velocity, and size within the target frame. Tracking target q i The state in frame t is represented as follows:

[0060]

[0061] where i∈N t N t It is the number of filters for the mismatched tracking target set UMT in frame t. Representing the target q i Location coordinates, Indicates the target size. This represents the average velocity of the three frames before the image was occluded.

[0062] During target movement, the target's geometric transformation characteristics can be used to predict its position and trajectory, thereby eliminating drift during tracking. By collecting target motion information and combining it with a weighted compensation method for re-matching, that is, by correcting and matching the target's position based on its motion trajectory information, tracking errors can be minimized.

[0063] The principle of this invention is as follows: First, the YOLO algorithm is used to detect targets in video footage from a substation. YOLO, as a real-time target detection algorithm, is characterized by high accuracy and speed. It transforms the target detection problem into a regression problem by dividing the image into multiple grid cells, each predicting a series of bounding boxes and corresponding class probabilities, thereby enabling the detection of multiple moving targets in the substation. Next, an Enhanced Kernel Correlation Filter (AKCF) is used for target tracking, utilizing the target's appearance information and features of the surrounding area. AKCF achieves accurate target tracking by minimizing the distance between the target features and candidate region features. This method effectively handles target occlusion issues, improving tracking accuracy and robustness. To further improve the accuracy of target tracking, a two-stage target association method is employed: In the first stage, tracking results are filtered and sorted using priority matching. Tracking results are sorted according to target priority, prioritizing the tracking results of high-priority targets to ensure accurate tracking of key targets. In the second stage, a motion estimation-based re-matching method is used to further correct and match the targets. By collecting target motion information and performing re-matching using a weighted compensation method—that is, correcting and matching the target's position based on its motion trajectory information—tracking errors can be minimized. This invention can play an important role in the safety monitoring and defect detection of smart substations, improving substation safety and operational efficiency.

[0064] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

[0065] The system and method provided in this embodiment can be stored in a computer-readable storage medium in the form of code, implemented as a computer program, and the basic parameter information required for calculation can be input through computer hardware, and the calculation results can be output.

[0066] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0067] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0068] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0069] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

[0071] This patent is not limited to the above-described preferred embodiments. Anyone can derive other forms of multi-target tracking method for on-site monitoring of intelligent substations based on the YOLO algorithm under the guidance of this patent. All equivalent changes and modifications made within the scope of this patent application shall fall within the scope of this patent.

Claims

1. A multi-target tracking method for on-site monitoring of intelligent substations based on the YOLO algorithm, characterized in that, Includes the following steps: Step S1: Use the YOLO algorithm to perform target detection on the video of the substation site, including dividing the image into multiple grid cells and performing regression on each cell to predict the bounding box and class probability, so as to achieve accurate detection of multiple moving targets in the substation site; Step S2: Target tracking is performed using the enhanced kernel correlation filter AKCF, where AKCF is a sample-trained filter that achieves accurate target tracking by minimizing the distance between target features and candidate region features. Step S3: A two-stage target association method based on priority matching and motion estimation rematch is adopted to increase tracking accuracy; Step S4: Use priority matching to filter and sort the tracking results. Sort the tracking results according to the priority of the target and retain the tracking results of high priority targets. Step S5: Correct and match the target using a motion estimation-based rematching method. Correct and match the target's position based on its motion trajectory information to reduce tracking error and improve tracking accuracy. In step S5, the goal of object association is to achieve object re-matching after long-term occlusion; the target's state is represented by the target frame's position, average velocity, and size; tracking target q i The state in frame t is represented as follows: where i∈N t N t It is the number of filters for the mismatched tracking target set UMT in frame t. Representing the target q i Location coordinates, Indicates the target size. This represents the average velocity of the three frames before the occlusion. During the target's movement, the geometric transformation characteristics of the target are used to predict the target's position and trajectory, thereby eliminating the drift phenomenon during the tracking process. By collecting the target's motion information and combining it with a weighted compensation method for re-matching, that is, the target's position is corrected and matched according to the target's motion trajectory information, so as to minimize the tracking error.

2. The multi-target tracking method for on-site monitoring of intelligent substations based on the YOLO algorithm according to claim 1, characterized in that: In step S4, it is assumed that in frame t, the M detection results represent the set of detected targets. Furthermore, the tracker's prediction results for N targets are represented as a set of tracked targets. ; definition Its UMT is a set of mismatched tracked targets, MP is a set of matched pairs; pv represents the maximum response value predicted by AKCF; IOU represents the degree of overlap between the tracked target frame and the detected target frame, and the calculation formula is as follows: Where S t S represents the target frame being tracked. d Indicates the target frame to be detected; The two-dimensional overlap correlation matrix IM is used to represent the similarity between targets: Among them, IOU NM Represents the Nth tracking target q in the tracking set TRA N With the Mth detection target p in the detection target set DET M The overlap of target frames between them.

3. A multi-target tracking system for intelligent substation field monitoring based on the YOLO algorithm, characterized in that: The method for performing multi-target tracking in intelligent substation field monitoring based on the YOLO algorithm as described in claim 1 includes the following modules: Grid cells used to divide video images from substation sites; Module used for regressing target bounding boxes and class probabilities; This module is used to calculate the distance between target features and candidate region features; This module is used to select high-priority targets for tracking. A module used for correcting target position and matching.

Images