A substation based on beidou positioning method and system for detecting regional intrusion

Abstract

The application relates to the technical field of substation security and protection, and particularly discloses a substation regional intrusion detection method and system based on Beidou positioning. An improved regional intrusion detection algorithm based on YOLOv5 is proposed, the algorithm takes Beidou accurate positioning information as the data input basis, takes the YOLOv5 network as the core target detection algorithm, introduces the CBAM attention mechanism in the backbone network, and adds an additional small target detection layer at the neck, effectively improving the feature extraction capability of the network and the detection accuracy of small target objects. In the detection part, the image pixel coordinate system and the mask parameter are applied to divide the alert area for intrusion detection. The Beidou positioning system and the UWB positioning system are fused, which can ensure global positioning in a wide geographical area and high-precision positioning requirements in complex indoor environments. The high-speed transmission capability of the 5G network is used to ensure the timely uploading and processing of data, so as to ensure the real-time performance and stability of the system.

Description

Technical Field

[0001] This invention relates to the field of substation security technology, and specifically discloses a method and system for area intrusion detection in substations based on BeiDou positioning. Background Technology

[0002] Substations are a crucial component of the power system, and their safety is closely linked to social development. Currently, substation operations mainly rely on regular inspections and maintenance by patrol personnel to assess equipment for defects and potential hazards. However, due to the inherently dangerous nature of substation operations, personnel safety is paramount. Traditional substation operation methods suffer from insufficient technological safeguards and a lack of scientific and efficient control measures. This can lead to delayed safety warnings for operators, preventing managers from real-time monitoring of the operation process, replaying operational data, or conducting real-time command and dispatch. This situation, to some extent, increases the risk of personal injury and property damage.

[0003] To address the aforementioned issues, this project faces three main challenges. First, the complex operating environment of substations, with severe electromagnetic interference, leads to inaccurate positioning equipment. Second, the project needs to solve the problem of integrated indoor and outdoor positioning. Traditional positioning methods are relatively singular and cannot achieve indoor and outdoor positioning functions on a single device. Finally, the positioning equipment requires high accuracy and real-time early warning capabilities to respond to intrusions by personnel within the area. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for area intrusion detection in substations based on BeiDou positioning, thereby improving the safety and efficiency of substation operations. This solution should consider the complexity of the substation operating environment and the impact of electromagnetic interference to ensure the accuracy of the positioning equipment. Simultaneously, it should overcome the challenges of indoor and outdoor positioning and possess real-time early warning and alarm functions to meet the requirements for the accuracy and safety of the positioning equipment. By addressing these pain points, we can improve the efficiency and safety of substation operations, reduce the risk of personnel casualties and property losses, and make a positive contribution to the stable operation of the power system and social development.

[0005] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0006] Firstly, a method for area intrusion detection in substations based on BeiDou positioning is provided, which specifically includes the following steps:

[0007] By combining the BeiDou positioning system to collect high-precision maps of substations, and based on the substation's safety zoning, the system application server presets the range of prohibited areas on the collected high-precision maps.

[0008] Several binocular cameras are installed around the substation or at key locations to collect image data in real time. The video management module is used to delineate the warning area and the safe area for the collected image data.

[0009] The binocular cameras are connected to the system application server via a 5G network to upload the collected image data in real time.

[0010] The target detection module uses an improved version of the single-stage target detection algorithm Yolov5 to detect pedestrian positions in the warning area of ​​real-time acquired image data and mark their positions.

[0011] Using BeiDou high-precision smart wearable devices, the precise location information of pedestrians can be obtained and uploaded in real time via 5G network;

[0012] When a pedestrian's location in the image data is detected to be within the warning area, the intrusion status judgment module sends a pedestrian location acquisition request, retrieves the pedestrian's location information in the corresponding image data, and compares the pedestrian's precise location information with the preset intrusion area range; if the pedestrian's location is within the preset intrusion area, it is determined that the pedestrian has intruded into the prohibited area, triggering the corresponding alarm and issuing alarm information;

[0013] When a pedestrian's location is detected within a preset prohibited area, the intrusion status judgment module sends a video acquisition request and uses the image data of the corresponding location for target detection. If the pedestrian is confirmed to be within the warning area, the corresponding alarm is triggered and an alarm message is issued.

[0014] The above technical solutions can achieve the following beneficial effects: accurate pre-setting of prohibited areas through high-precision map collection; timely data upload and processing by utilizing the high-speed transmission capabilities of 5G networks; improved accuracy and real-time performance of target detection by adopting an improved version of the single-stage target detection algorithm Yolov5; and accuracy of intrusion detection and reduction of false alarms and missed alarms by dual verification of target detection and accurate location information.

[0015] Preferably, the method for delineating the warning area and safe area of ​​the collected image data through the video management module includes: extracting the four-point pixel coordinates of the bounding rectangle of the area to be detected in the pixel coordinate system of the image based on the image location information and image pixel coordinates provided by the Beidou radar, and forming the detection area, i.e. the warning area, by combining the mask parameters. When performing target detection, only the warning area is detected, which reduces the detection range and improves the detection efficiency.

[0016] Preferably, the warning zone includes at least a prohibited area to prevent missed detection.

[0017] Preferably, the improved Yolov5 algorithm includes: introducing the CBAM attention mechanism into the Yolov5 backbone network, which increases the weight of small targets in the whole feature map, making it easier for the network to learn the feature information of the target to be detected, while avoiding the feature information of the target to be detected being masked by redundant background information during convolution, thereby improving detection accuracy, saving parameters and computing power, and improving the accuracy of model training.

[0018] Preferably, the improved Yolov5 algorithm further includes: adding an additional small target detection layer in the Prediction part, concatenating the shallower feature map with the deeper feature map before detection; thus enhancing the network's ability to detect small targets.

[0019] Preferably, the target detection module utilizes an improved version of the single-stage target detection algorithm YOLOv5 to detect pedestrian positions in the warning area within the real-time acquired image data and mark their locations. In addition, it also includes: video analysis and early warning of the real-time acquired image data; using video analysis algorithms to process and analyze the acquired image data; and by recognizing pedestrian actions and behavioral patterns, combined with historical data and rule settings, providing early warnings of potential safety risks so that relevant personnel can take necessary measures to prevent accidents.

[0020] Preferably, the Beidou high-precision smart wearable device can simultaneously support positioning using both the Beidou positioning system and the UWB positioning system. When a pedestrian is outdoors, it mainly relies on the Beidou positioning system for positioning. When the pedestrian enters indoors or the coverage area of ​​a UWB base station, it automatically switches to the UWB positioning system for more accurate positioning, thus solving the problem of accurate positioning both indoors and outdoors.

[0021] Secondly, a BeiDou-based area intrusion detection system for substations is provided. This system includes: a system application server, several binocular cameras, and a BeiDou high-precision smart wearable device. The system application server is used for high-precision map management of the substation, pre-setting prohibited areas; receiving image data collected by the binocular cameras, delineating warning areas, and performing target detection on the image information; receiving precise pedestrian location information uploaded by the BeiDou high-precision smart wearable device; judging the intrusion status and triggering an alarm, issuing alarm information; and simultaneously performing video analysis and early warning on the real-time collected image data. The binocular cameras are used to collect image data of the substation's surroundings and key locations in real time, providing image location information; the BeiDou high-precision smart wearable device is used to obtain the precise location information of pedestrians entering the substation.

[0022] Preferably, the system application server includes: a data receiving module for receiving image data collected by the plurality of binocular cameras and receiving pedestrian precise location information uploaded by the Beidou high-precision smart wearable device; a video management module for managing the received image data and manually or automatically delineating the warning area and safe area of ​​the collected image data according to a preset prohibited area; a target detection module for executing an improved algorithm of the single-stage target detection algorithm YOLOv5 to detect pedestrian positions in the warning area of ​​the real-time collected image data and mark their positions; and simultaneously analyzing and issuing warnings for the real-time collected image data; an intrusion judgment module for performing dual verification based on the detection results of the target detection module and the comparison results between the pedestrian position information and the preset prohibited area to determine the pedestrian intrusion status; and an early warning module for triggering an alarm and issuing alarm information when the intrusion judgment module confirms that a pedestrian has intruded into the prohibited area; or issuing an early warning when the target detection module detects abnormal actions or behaviors of the pedestrian.

[0023] Preferably, the plurality of binocular cameras and the Beidou high-precision smart wearable device transmit data with the system application server via a 5G network.

[0024] The beneficial effects of the technical solution provided by the embodiments of the present invention are as follows:

[0025] This invention addresses the problems of high installation and maintenance costs, high false alarm rates, and significant security risks associated with traditional area intrusion detection technologies. It proposes a BeiDou-based area intrusion detection algorithm for substations. The algorithm uses precise BeiDou positioning information as its data input and employs a YOLOv5 network as its core target detection algorithm. A CBAM attention mechanism is introduced into the backbone network, and an additional small target detection layer is added at the neck layer, effectively improving the network's feature extraction capabilities and enhancing the detection accuracy for small targets. In the detection section, image pixel coordinates and mask parameters are used to divide the warning area for intrusion detection. Experiments show that the proposed algorithm has good real-time performance, adapts to environmental changes, and achieves high accuracy in intrusion detection. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, without exceeding the scope of protection claimed in this application.

[0027] Figure 1 This is a flowchart of a substation area intrusion detection method based on BeiDou positioning provided in an embodiment of the present invention.

[0028] Figure 2 This is a schematic diagram of a process for forming a detection area using a mask method, provided by an embodiment of the present invention.

[0029] Figure 3 This is a schematic diagram of a network architecture for an improved Yolov5 algorithm provided in an embodiment of the present invention.

[0030] Figure 4 This is a schematic diagram of a model training process after introducing the CBAM attention mechanism, as provided in an embodiment of the present invention.

[0031] Figure 5 This is a schematic diagram of the network Neck and Prediction structure after adding a small target detection layer according to an embodiment of the present invention.

[0032] Figure 6 This is a schematic diagram of the system structure of a substation area intrusion detection system based on BeiDou positioning provided in an embodiment of the present invention.

[0033] Figure 7 This is a schematic diagram of the system structure of a substation area intrusion detection system based on BeiDou positioning provided in an embodiment of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages 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 of the invention and are not intended to limit the invention. It should be noted that step designations such as S101 and S102 are used in this description for the purpose of more clearly and concisely describing the corresponding content, and do not constitute a substantial limitation on the order of steps.

[0035] Figure 1 This is a flowchart of a substation area intrusion detection method based on BeiDou positioning provided in an embodiment of the present invention.

[0036] Specifically, the following steps are included:

[0037] Step S1: Combine the BeiDou positioning system to collect high-precision maps of the substation. Based on the substation's safety zoning, the system application server presets the range of prohibited areas on the collected high-precision map.

[0038] Combining the BeiDou positioning system with high-precision map collection of substations can enable precise pre-setting of prohibited areas and improve positioning accuracy.

[0039] Step S2: Install several binocular cameras around the substation or at key locations, collect image data in real time through the binocular cameras, and delineate the warning area and safe area of ​​the collected image data through the video management module.

[0040] By acquiring image data through binocular cameras, the depth information of objects can be calculated from the differences between the images captured by the two lenses, enabling three-dimensional spatial perception. This helps to accurately delineate the boundaries of the warning area and accurately identify targets entering the warning area.

[0041] Step S3: The plurality of binocular cameras are connected to the system application server via a 5G network to upload the collected image data in real time;

[0042] Leveraging the high-speed transmission capabilities of 5G networks ensures timely data upload and processing.

[0043] Step S4: The target detection module uses an improved algorithm of the single-stage target detection algorithm Yolov5 to detect pedestrian positions in the warning area in the real-time acquired image data and marks their positions.

[0044] Step S5: Use BeiDou high-precision smart wearable devices to obtain accurate pedestrian location information and upload the accurate pedestrian location information in real time via 5G network;

[0045] Step S6: When the pedestrian's location in the image data is detected to be within the warning area, the intrusion status judgment module sends a pedestrian location acquisition request, calls the pedestrian's location information in the corresponding image data, and compares the pedestrian's precise location information with the preset intrusion area range; if the pedestrian's location is within the preset intrusion area, it is determined that the pedestrian has intruded into the prohibited area, triggers the corresponding alarm, and issues an alarm message.

[0046] Step S7: When the pedestrian's location information is detected within the preset prohibited area, the intrusion status judgment module sends a video acquisition request and calls the image data of the corresponding location for target detection; if it is confirmed that the pedestrian is within the warning area, the corresponding alarm is triggered and an alarm message is issued.

[0047] It should be noted that by combining BeiDou positioning data and video target detection results, the system can accurately determine the specific location information of each detected target. This dual verification mechanism improves the accuracy and reliability of regional intrusion detection, prevents false alarms and missed alarms, and ensures timely and effective security protection.

[0048] In some other embodiments of this application, the method of delineating the warning area and safe area of ​​the collected image data through the video management module includes: extracting the four-point pixel coordinates of the bounding rectangle of the area to be detected in the pixel coordinate system of the image based on the image location information and image pixel coordinates given by the Beidou radar, and forming the detection area, i.e. the warning area, by combining the mask parameters. When performing target detection, only the warning area is detected, which reduces the detection range and improves the detection efficiency.

[0049] It should be noted that the application of a mask can be understood as a binary image composed of 0s and 1s. After masking the area to be detected, areas with a value of 1 are displayed normally, while areas with a value of 0 are masked and ignored during target detection. This allows the masking method to form the detection area, also known as the warning area. The main process is as follows: Figure 2 As shown: Figure 2 (b) is a binary image composed of 0s and 1s; Figure 2 (c) is the mask image in the intrusion detection scenario in this paper, where the red area is the warning area (the shape is determined by the coordinates of the four points that make up the warning area, and can be flexibly adjusted according to the actual situation), and the black area is the shielding area; Figure 2 (d) is a combination of the mask image and the background image, clearly showing that the area outside the warning zone is completely blocked. Specifically, the principle of the intrusion detection method in this paper can be understood as follows: if a target is detected within the warning zone, it is considered that "the target has intruded into the area"; if no target is detected, no intrusion judgment is made.

[0050] In some other embodiments of this application, the warning area includes at least a prohibited area to prevent missed detection.

[0051] It should be noted that including prohibited areas within the warning zone helps the system detect pedestrians approaching prohibited areas in advance, enabling real-time tracking, early warning, and preventing missed detections.

[0052] In other embodiments of this application, the improved Yolov5 algorithm includes: introducing a CBAM attention mechanism into the Yolov5 backbone network, which increases the weight of small targets in the entire feature map, making it easier for the network to learn the feature information of the target to be detected, while avoiding the feature information of the target to be detected being masked by redundant background information during convolution, thereby improving detection accuracy, saving parameters and computing power, and improving the accuracy of model training. Its network architecture is as follows: Figure 3 As shown.

[0053] It should be noted that when in an outdoor environment, it is necessary to extract the feature information of abnormal targets from complex background information and filter out irrelevant information. When there are many small target data in the custom dataset, because small targets occupy few pixels, they are prone to losing feature information in deep networks, leading to missed detections and false detections, affecting the overall detection accuracy of the model. Therefore, the CBAM attention mechanism is introduced into the backbone network. By increasing the weight of small targets in the entire feature map, the feature information of the target to be detected is made easier for the network to learn, while avoiding the target's feature information being masked by redundant background information during convolution. The CBAM attention mechanism is mainly divided into a channel attention module and a spatial attention module. The channel attention mechanism focuses on detecting the content of the target, while the spatial attention mechanism focuses on detecting the position of the target. The combination of the two can make the main features of the target more focused, effectively saving parameters and computing power. At the same time, the CBAM attention mechanism module can suppress the interference of general features, thereby improving the accuracy of model training. The specific process is shown in Figure 4.

[0054] The corresponding process expression is:

[0055]

[0056] In some other embodiments of this application, the improved Yolov5 algorithm further includes: adding an additional small target detection layer in the Prediction part, concatenating the shallower feature map with the deeper feature map before detection; thus enhancing the network's ability to detect small targets.

[0057] It should be noted that the main reason why YOLOv5 performs poorly in detecting small targets is its large downsampling factor, which makes it difficult for the deeper feature maps to learn the feature information of small targets. Therefore, some embodiments choose to add a small target detection layer, concatenating the shallower feature map with the deeper feature map before detection. The original YOLOv5 network model has only three detection layers, corresponding to three sets of initial anchor values, where small-value anchors detect large targets, and large-value anchors detect small targets. Now, based on the existing foundation, an additional set of Anchors ([5,6; 8,15; 16,11]) is added, with values ​​approximately half that of the smallest Anchor set. Then, seven operation layers are added after layer 17 to further upsample the feature map, expanding it further. Simultaneously, in layer 20 of the improved network, the acquired feature map is concat-fused with the feature map from layer 2 of the backbone network to obtain a larger feature map for small object detection. A small object detection layer [12-13] is added after layer 31, bringing the total to four layers for detection, enhancing the network's ability to detect small objects. After adding the small object detection layer, the network's Neck and Prediction structures are as follows: Figure 5 As shown.

[0058] In other embodiments of this application, the target detection module utilizes an improved algorithm of the single-stage target detection algorithm YOLOv5 to detect pedestrian positions in the warning area of ​​the real-time acquired image data and mark their positions. In addition, it also includes: video analysis and early warning of the real-time acquired image data; using video analysis algorithms to process and analyze the acquired image data; and by recognizing pedestrian actions and behavioral patterns, combined with historical data and rule settings, providing early warnings of potential safety risks so that relevant personnel can take necessary measures to avoid safety accidents.

[0059] In some other embodiments of this application, the Beidou high-precision smart wearable device can simultaneously support positioning using both the Beidou positioning system and the UWB positioning system. When a pedestrian is outdoors, the device primarily relies on the Beidou positioning system for positioning. When the pedestrian enters indoors or within the coverage area of ​​a UWB base station, the device automatically switches to the UWB positioning system for more accurate positioning, thus solving the problem of accurate positioning both indoors and outdoors.

[0060] It should be noted that the solution employs a fusion of the BeiDou and UWB positioning systems. When the target is outdoors, it primarily relies on the BeiDou system for positioning. When the target enters indoors or within the coverage area of ​​a UWB base station, it automatically switches to the UWB system for more precise positioning. Indoor and outdoor positioning data can be transmitted in real time to the system application server via 4G / 5G networks, local area networks, etc. The server processes the received data and outputs continuous and accurate positioning information. The entire system can guarantee global positioning over a wide geographical area while also ensuring high-precision positioning in complex indoor environments.

[0061] Figure 6 This is a schematic diagram of the system structure of a substation area intrusion detection system based on BeiDou positioning provided in an embodiment of the present invention.

[0062] The system includes: a system application server 601, several binocular cameras 602, and a Beidou high-precision smart wearable device 603. The system application server 601 is used for high-precision map management of the substation, pre-setting prohibited areas; receiving image data collected by the binocular cameras, delineating warning areas, and performing target detection on the image information; receiving precise pedestrian location information uploaded by the Beidou high-precision smart wearable device; determining intrusion status and triggering alarms; and simultaneously performing video analysis and early warning on the real-time collected image data. The binocular cameras 602 are used to collect image data of the substation's surroundings and key locations in real time, providing image location information. The Beidou high-precision smart wearable device 603 is used to obtain precise location information of pedestrians entering the substation.

[0063] Furthermore, the plurality of binocular cameras and the Beidou high-precision smart wearable device transmit data with the system application server via a 5G network.

[0064] Figure 7 This is a schematic diagram of the system structure of a substation area intrusion detection system based on BeiDou positioning provided in an embodiment of the present invention.

[0065] The system application server includes: a data receiving module 701, used to receive image data collected by the plurality of binocular cameras and to receive pedestrian precise location information uploaded by the Beidou high-precision smart wearable device; a video management module 702, used to manage the received image data and manually or automatically delineate the warning area and safe area of ​​the collected image data according to the preset prohibited area; a target detection module 703, used to execute the improved algorithm of the single-stage target detection algorithm YOLOv5, to detect pedestrian positions in the warning area of ​​the real-time collected image data and mark their positions; and to analyze and issue warnings for the real-time collected image data; an intrusion judgment module 704, used to perform dual verification based on the detection results of the target detection module and the comparison results of the pedestrian position information with the preset prohibited area to judge the pedestrian intrusion status; and an early warning module 705, used to trigger an alarm and issue an alarm message when the intrusion judgment module confirms that a pedestrian has intruded into the prohibited area; or to issue an early warning when the target detection module detects abnormal actions or behaviors of the pedestrian.

[0066] In summary, this invention addresses the problems of high installation and maintenance costs, high false alarm rates, and significant security risks associated with traditional regional intrusion detection technologies by providing a substation regional intrusion detection method and system based on BeiDou positioning. An improved regional intrusion detection algorithm based on YOLOv5 is proposed. This algorithm uses precise BeiDou positioning information as data input and the YOLOv5 network as the core target detection algorithm. A CBAM attention mechanism is introduced into the backbone network, and an additional small target detection layer is added at the neck, effectively improving the network's feature extraction capabilities and increasing the detection accuracy for small targets. In the detection part, the warning area is divided using image pixel coordinates and mask parameters for intrusion detection. A fusion scheme of BeiDou and UWB positioning systems is adopted, ensuring global positioning over a wide geographical area while maintaining high-precision positioning in complex indoor environments. The high-speed transmission capability of the 5G network ensures timely data upload and processing, guaranteeing the system's real-time performance and stability. Experiments show that the proposed method has good real-time performance, adaptability to environmental changes, and high accuracy in intrusion detection.

Claims

1. A regional intrusion detection method based on Beidou positioning of a substation, characterized in that, The application relates to a high-precision map acquisition system for a substation based on a Beidou positioning system. The system comprises the following steps: A plurality of binocular cameras are arranged around the substation or at key positions, and the binocular cameras are used to collect image data in real time; A video management module extracts four pixel coordinates of an external rectangle of a detection area according to image position information and image pixel coordinates provided by a Beidou radar, forms a detection area by combining mask parameters, and defines the detection area as a warning area, wherein the warning area at least contains the forbidden area; The binocular cameras are connected to a system application server through a 5G network and upload the collected image data in real time; A target detection module uses an improved algorithm of a single-stage target detection algorithm Yolov5 to detect the position of a pedestrian in the warning area of the real-time collected image data and mark the position of the pedestrian; The improved Yolov5 target detection algorithm introduces a CBAM attention mechanism into a backbone network of the Yolov5 network and adds an additional small target detection layer in a prediction part of the Yolov5 network; A Beidou high-precision intelligent wearable device is used to obtain accurate position information of the pedestrian, wherein the Beidou high-precision intelligent wearable device supports Beidou positioning system and UWB positioning system for positioning at the same time; when the pedestrian is outdoors, the Beidou positioning system is used for positioning, and when the pedestrian enters indoors or a UWB base station coverage range, the UWB positioning system is automatically switched to for positioning; and the accurate position information of the pedestrian is uploaded in real time through a 5G network; When the target detection module detects that the position of the pedestrian in the image data is in the warning area, an intrusion state judgment module sends a pedestrian position acquisition request, calls the position information of the pedestrian in the corresponding image data, and compares the accurate position information of the pedestrian with a preset forbidden area range; If the position of the pedestrian is in the preset forbidden area, it is judged that the pedestrian intrudes into the forbidden area, a corresponding alarm is triggered, and alarm information is sent. When the accurate position information of the pedestrian is detected in the preset forbidden area, the intrusion state judgment module sends a video acquisition request, calls the image data of the corresponding position, and performs target detection by the target detection module; if it is confirmed that the pedestrian is in the warning area in the image data, it is judged that the pedestrian intrudes into the forbidden area, a corresponding alarm is triggered, and alarm information is sent.

2. The regional intrusion detection method based on Beidou positioning of a substation according to claim 1, characterized in that: The The CBAM attention mechanism is introduced into the Yolov5 backbone network by increasing the weight of small targets in the whole feature map, so that the feature information of the to-be-detected target is more easily learned by the network, and the feature information of the to-be-detected target is prevented from being covered by redundant background information during convolution. The target detection module only detects the alert area when performing target detection.

3. The regional intrusion detection method based on Beidou positioning of a substation according to claim 1, characterized in that, The additional small target detection layer is added in the prediction part, and the relatively shallow feature map and the relatively deep feature map are spliced before detection.

4. The regional intrusion detection method based on Beidou positioning of a substation according to claim 1, characterized in that, The target detection module uses the improved algorithm of the single-stage target detection algorithm Yolov5 to detect the position of the pedestrian in the warning area of the real-time collected image data and mark the position of the pedestrian.

5. The regional intrusion detection method based on Beidou positioning of a substation according to claim 1, characterized in that, ​ Video analysis and early warning on real-time collected image data: using video analysis algorithm to process and analyze the collected image data; through identifying the action and behavior pattern of the pedestrian, combining historical data and rule setting, early warning potential safety risks.

6. A substation regional intrusion detection system based on Beidou positioning, characterized in that, It comprises a system application server, a plurality of binocular cameras, and a Beidou high-precision intelligent wearable device, wherein: The system application server is used for high-precision map management of the substation in combination with the Beidou positioning system and presetting of prohibited areas; receives image data collected by the plurality of binocular cameras, divides a warning area according to a mask parameter through a video management module, the warning area at least containing the prohibited area; uses a target detection module to perform target detection on image information and video analysis and early warning on real-time collected image data; the target detection module executes an improved YOLOv5 target detection algorithm, which introduces a CBAM attention mechanism in the backbone network of the YOLOv5 network and adds an additional small target detection layer in the prediction part; receives pedestrian accurate position information uploaded by the Beidou high-precision intelligent wearable device; performs intrusion state judgment through an intrusion judgment module and triggers an alarm to issue an alarm information; The plurality of binocular cameras are arranged around the substation or at key positions and are used for real-time collection of image data; The Beidou high-precision intelligent wearable device is used for obtaining accurate position information of pedestrians entering the substation and can simultaneously support the Beidou positioning system and the UWB positioning system for positioning; when the pedestrian is outdoors, the Beidou positioning system is relied on for positioning, and when the pedestrian enters indoors or the UWB base station coverage range, the UWB positioning system is automatically switched to for positioning.

7. The regional intrusion detection system based on Beidou positioning of a substation according to claim 6, characterized in that, The system application server comprises: A data receiving module is used for receiving image data collected by the plurality of binocular cameras and receiving pedestrian accurate position information uploaded by the Beidou high-precision intelligent wearable device; A video management module is used for managing the received image data and manually or automatically dividing the warning area and the safety area of the collected image data according to the preset prohibited area; A target detection module is used for executing an improved algorithm of a single-stage target detection algorithm Yolov5, performing target detection on the position of the pedestrian in the warning area of the real-time collected image data, and marking the position; and simultaneously performing analysis and early warning on the real-time collected image data; An intrusion judgment module is used for double verification according to the detection result of the target detection module and the comparison result of the pedestrian position information and the preset prohibited area to judge the intrusion state of the pedestrian; An early warning module is used for triggering an alarm to issue an alarm information when the intrusion judgment module confirms that the pedestrian intrudes into the prohibited area or the target detection module detects that the pedestrian has abnormal action or behavior to issue an early warning.

8. The substation regional intrusion detection system based on the Beidou positioning according to claim 7, wherein: The plurality of binocular cameras and the Beidou high-precision intelligent wearable device perform data transmission with the system application server through a 5G network.

Images