Catenary fault identification method based on video monitoring
By deploying video surveillance equipment in railway stations and along the railway line, and combining it with multi-dimensional databases and video image anomaly recognition algorithms, the problems of long detection cycles and delayed responses of overhead contact line faults have been solved, enabling real-time monitoring and rapid response to overhead contact line faults, and improving the timeliness and accuracy of fault identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-21
- Publication Date
- 2026-03-03
AI Technical Summary
Existing technologies for detecting catenary faults suffer from problems such as long detection cycles, delayed responses, and insufficient accuracy, making it difficult to meet the high reliability requirements of modern railways for the safe operation of catenary systems.
A video surveillance-based contact network fault identification method is adopted. This method utilizes existing railway stations and along the railway line video surveillance equipment, combined with a multi-dimensional database and video image anomaly recognition algorithm, to monitor the contact network status in real time. Suspected faults are identified through manual confirmation, and a multi-dimensional correlation between the line, station, track, power supply arm, and substation is constructed to achieve rapid fault location and accurate analysis.
It enables real-time monitoring and rapid response to overhead contact line faults, significantly improving the efficiency of fault detection and the timeliness of fault identification, shortening fault handling time, and enhancing the operational reliability of the overhead contact line system.
Smart Images
Figure CN121600440A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power supply safety monitoring technology for electrified railways, specifically to a method for identifying overhead contact line faults based on video surveillance. Background Technology
[0002] As the core power supply system of electrified railways, the overhead contact system's operational status directly affects the normal operation and safety of trains. Erected above the railway line, this system uses high-voltage power transmission to power electric locomotives. However, exposed to a complex and ever-changing external environment, it is susceptible to interference from various factors. For example, treetops growing along the line may encroach on the contact system's clearance; kites, mulch film, and other floating objects may become entangled in the contact wires; and birds nesting, moving about, or snakes coming into contact with live parts can all trigger discharge faults or short-circuit trips. These faults are characterized by their suddenness and wide-ranging impact; failure to detect and handle them promptly can lead to train stoppages, equipment damage, and even safety accidents. Furthermore, the environment in which the contact system is located is also subject to natural factors such as wind and sand erosion, snow cover, and lightning strikes, further increasing the difficulty of operation and maintenance.
[0003] Currently, the detection of overhead contact line faults mainly relies on two technical means. The first is the overhead contact line safety inspection device, also known as a 2C device, which performs image acquisition and analysis of the line periodically (usually every 10 days) to achieve preventative detection of potential hazards. However, this method is limited by the inspection frequency and cannot meet the real-time requirements of environmental changes, especially in extreme weather or sudden foreign object intrusion, resulting in detection blind spots. The second is the fault locator, which calculates the fault location by analyzing electrical parameters at the time of tripping, assisting maintenance personnel in subsequent troubleshooting. However, this method only provides approximate mileage information, still requiring manual on-site inspection to determine the specific cause of the fault, leading to low response efficiency. Furthermore, because electrified railway lines are generally hundreds of kilometers long, limitations in funding and resource investment currently prevent the large-scale deployment of cameras or sensors along the entire line, further exacerbating the lag and passivity of fault detection.
[0004] Therefore, existing technologies have significant shortcomings in terms of timeliness, accuracy, and economy, making it difficult to meet the high reliability requirements of modern railways for the safe operation of overhead contact lines. With the continuous expansion of railway networks and the increase in operational density, traditional detection methods are no longer adequate for the increasingly complex maintenance needs, necessitating a technological solution for all-weather, high-precision, and low-cost overhead contact line fault monitoring. Summary of the Invention
[0005] Based on the current state of the technology, the purpose of this invention is to solve the problems of long detection cycles and time-consuming and laborious cause investigation in overhead contact lines. Therefore, a video surveillance-based method for identifying overhead contact line faults is proposed. This method utilizes video surveillance cameras installed in existing railway stations and key locations along the line. It extracts video surveillance recordings from a constructed multi-dimensional database of these cameras, uses a video image anomaly recognition algorithm to extract intrusion and discharge features, outputs and returns suspected overhead contact line faults, and allows for manual confirmation of the identification results. This invention can promptly identify and respond to potential safety hazards and faults in the overhead contact line, shorten fault handling time, and improve fault handling efficiency.
[0006] The present invention employs the following technical solutions to achieve its objective: A method for identifying overhead contact line faults based on video surveillance, the method comprising the following steps: S1. Initialize the parameters of the multidimensional database, which includes information on video surveillance equipment, video image anomaly recognition algorithms, and railway line information. S2. After parameter initialization is complete, extract the video images already collected by the specified video surveillance equipment using either timed extraction or manual extraction when a contact network fault occurs. S3. Push the extracted video images to the video image anomaly recognition algorithm that has completed parameter initialization for video image analysis; S4. After completing the video image analysis, the video image anomaly recognition algorithm outputs images of suspected contact network faults and corresponding fault type results. S5. Based on the output of the video image anomaly recognition algorithm, suspected faults in the overhead contact line are confirmed manually.
[0007] Preferably, in step S1, the parameters of the railway line information are first initialized, and then the video monitoring equipment information and video image anomaly recognition algorithm information are configured, thereby completing the parameter initialization process of the multidimensional database as a whole. The railway line information includes the line, the section station, the track, the substation and the power supply arm. After collecting the basic data of these five types of objects, the data is entered item by item according to the logical relationship between the basic data of each type.
[0008] Furthermore, the logical relationships of the basic data of the five types of objects—line, section station, track, substation, and power outage supply arm—are as follows: track → section station → line, that is, track is associated with section station, and section station is associated with line; substation → line, that is, substation is associated with line; track ← power outage supply arm → substation, that is, power outage supply arm is associated with both substation and track.
[0009] Preferably, the video surveillance equipment uses cameras, and the basic information and multi-dimensional data relationships of the cameras are configured. The basic information includes camera name, camera ID, service IP, port number and location kilometer marker. The multi-dimensional data relationships include the lines, stations and tracks associated with the camera. The video image anomaly recognition algorithm includes multiple sub-algorithm recognition points, and each sub-algorithm recognition point is configured with a corresponding algorithm name and algorithm number.
[0010] Specifically, in step S2, the timed extraction method is as follows: for multiple video surveillance devices arranged in sequence, a preset extraction interval is used. Once the most recent image extraction is completed, the most recent extraction time is recorded as... Then the next extraction time When the time arrives At that time, the video storage server interface is called, and the video surveillance device IDs are passed in sequentially according to their numbers. The corresponding video images are then extracted, and the process is reset after extraction. and To achieve based on extraction interval The timed extraction process is performed.
[0011] Preferably, in step S2, the manual extraction method when a contact wire fault occurs is as follows: a preset manual extraction time is set for the video surveillance equipment. When a contact network tripping fault occurs, the operator manually selects the faulty line, substation, and power outage arm, and inputs the tripping kilometer marker and tripping time calculated by the fault distance measuring instrument. Based on the logical relationship of the basic data of the five types of objects—line, section station, track, substation, and power outage arm—and the multi-dimensional data relationship pre-configured by the video monitoring equipment, a list of associated video monitoring equipment related to the contact network tripping fault is obtained.
[0012] Preferably, based on the video surveillance equipment association list, video surveillance equipment within a preset distance threshold is selected according to the trip kilometer marker and the location kilometer marker in the basic information of the video surveillance equipment. The equipment is then re-arranged into an extraction list based on its distance from the trip kilometer marker, from closest to furthest. The start time for video image extraction is then determined to be a preset duration before the trip time. Subsequently, the video storage server interface is called, and the video surveillance equipment ID information and extraction start time are passed in sequentially according to the order of the video surveillance equipment in the extraction list, and the corresponding extraction duration is set. The video image.
[0013] Specifically, in step S3, the extracted video images are automatically pushed to the video image anomaly recognition algorithm interface based on the HTTP protocol. After receiving the pushed video images, the algorithm interface calls the corresponding algorithms to perform video image analysis according to the configured sub-algorithm recognition points. The sub-algorithm recognition points may include bird nests, floating objects, and tree encroachment. The video image analysis includes image cutting, image preprocessing, image restoration, image compression, image segmentation, and image recognition and analysis.
[0014] Specifically, in step S4, the video image anomaly recognition algorithm outputs the results of the video image analysis and returns them to the user. The output of the suspected contact network fault includes the fault image, fault type, fault occurrence time and corresponding fault video frame number.
[0015] Specifically, in step S5, the video image anomaly recognition algorithm presents suspected fault information of the overhead contact line through its output. The user performs manual analysis and confirmation based on the suspected fault information of the overhead contact line. Data that is confirmed as a real fault by manual analysis is marked as valid. After supplementing the detailed description of the real fault and the handling response information, it is stored in the overhead contact line fault database for summary. Data that is confirmed as having no fault by manual analysis is marked as invalid and discarded.
[0016] In summary, due to the adoption of this technical solution, the beneficial effects of this invention are as follows: This invention introduces video surveillance image anomaly recognition technology to achieve real-time monitoring and rapid response to overhead contact line faults. Compared to traditional methods relying on fixed-period detection, this invention utilizes video surveillance equipment deployed in existing railway stations and key locations along the line to continuously collect dynamic image data of the overhead contact line's operating status. This effectively overcomes the problems of long detection cycles and slow response times in existing technologies, significantly improving the efficiency of detecting potential safety hazards and the timeliness of fault identification. Through intelligent analysis of key parts of the overhead contact line using video image anomaly recognition algorithms, typical fault characteristics such as foreign object suspension, equipment detachment, and component deformation can be captured in real time, providing accurate visual evidence support for subsequent fault handling.
[0017] This invention constructs a multi-dimensional database system that establishes multi-dimensional relationships between lines, stations, tracks, power supply arms, and substations, achieving systematic integration of catenary equipment information. This database can quickly locate relevant camera resources when a fault occurs, effectively narrowing the video retrieval range and significantly improving fault identification efficiency. By organically integrating spatial location information with electrical equipment parameters, it not only enhances the accuracy of fault location but also provides multi-dimensional data support for subsequent fault analysis, thereby shortening fault investigation time and improving the operational reliability of the catenary system.
[0018] This invention employs a five-level data topology modeling approach to construct a spatial-electrical composite relationship model encompassing lines, stations, tracks, power supply arms, and substations. This model forms a complete overhead contact line equipment topology network through the association between stations and lines, the binding of tracks to stations, the correspondence between substations and lines, and the bidirectional association between power supply arms and substations / tracks. Combined with precise mileage data provided by fault location instruments, this invention upgrades traditional two-dimensional fault location to a three-dimensional coordinate location system that includes line mileage, track location, and electrical segmentation, achieving precise mapping of image features, spatial location, and electrical parameters. For example, when the power supply arm of substation 213 at Golmud East trips due to a power outage, and the fault location instrument indicates a mileage of K815+426, the associated model automatically locks onto the camera at platform 3, westbound, at Golmud station.
[0019] This invention effectively solves the problem of the disconnect between fault location and electrical equipment information in the prior art through the above-mentioned multi-dimensional positioning method, making accurate identification of fault points possible and providing reliable technical support for rapid fault handling and power restoration. Attached Figure Description
[0020] Figure 1 This is a schematic diagram illustrating the overall process of the overhead contact line fault identification method of the present invention. Figure 2 This is a flowchart illustrating the manual extraction method for when a contact network fault occurs in the present invention. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0022] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0023] Example A method for identifying overhead contact line faults based on video surveillance, which can also be seen in [the following text is missing]. Figure 1 The simplified flow is shown below, and the core steps of this method are as follows: S1. Initialize the parameters of the multidimensional database, which includes information on video surveillance equipment, video image anomaly recognition algorithms, and railway line information. S2. After parameter initialization is complete, extract the video images already collected by the specified video surveillance equipment using either timed extraction or manual extraction when a contact network fault occurs. S3. Push the extracted video images to the video image anomaly recognition algorithm that has completed parameter initialization for video image analysis; S4. After completing the video image analysis, the video image anomaly recognition algorithm outputs images of suspected contact network faults and corresponding fault type results. S5. Based on the output of the video image anomaly recognition algorithm, suspected faults in the overhead contact line are confirmed manually.
[0024] The application of the method in this embodiment is based on the intelligent monitoring system for railway catenary. The details of each step will be described in detail and with preferred examples, following the order of the steps described above.
[0025] The purpose of step S1 is to construct a multidimensional database with topological relationship mapping capabilities as the core data carrier. This database adopts a hybrid architecture of relational database and graph database, in which railway line information storage uses a hierarchical modeling approach. For the parameter initialization process of the multidimensional database, the railway line information is first initialized, and then the video surveillance equipment information and video image anomaly recognition algorithm information are configured. In this embodiment, the railway line information includes lines, station yards, tracks, substations, and power outage supply arms. After collecting the basic data of these five types of objects, the data is entered item by item according to the logical relationships between each type of basic data.
[0026] The logical relationships of the basic data of the five types of objects selected in this embodiment—line, station yard, track, substation, and power outage supply arm—are as follows: Track → Station Yard → Line, that is, track is associated with station yard, and station yard is associated with line; Substation → Line, that is, substation is associated with line; Track ← Power Outage Supply Arm → Substation, that is, power outage supply arm is associated with both substation and track. Specific association examples are shown below in Table 1.
[0027] Table 1. Schematic diagram of logical relationships between railway line information
[0028] In this embodiment, the video surveillance equipment uses cameras, and the camera association can be seen in Table 1 above. The parameter configuration of the video surveillance equipment adopts a dual binding mechanism. When entering the basic information of the camera, in addition to the usual basic parameters such as device name, device ID, and network address, it is mandatory to bind the corresponding line topology node. Therefore, it is necessary to configure the basic information and multi-dimensional data relationships of the camera. The basic information includes the camera name, camera ID, service IP, port number, and location kilometer marker. The multi-dimensional data relationships include the line, section station, and track associated with the camera. The video image anomaly recognition algorithm includes multiple sub-algorithm recognition items, and each sub-algorithm recognition item is configured with a corresponding algorithm name and algorithm number. This multi-dimensional data association provides a key index foundation for subsequent fault location. When the system detects that a power supply arm has tripped, it can immediately retrieve the track served by that power supply arm and the corresponding monitoring equipment through the topology network.
[0029] Step S2 determines the video extraction trigger mechanism. In this embodiment, the system can provide a dual-mode collaborative operation of timed polling and event-driven methods, which can be manifested as timed extraction or manual extraction when a contact network fault occurs.
[0030] In the timed extraction method, for multiple video surveillance devices arranged in sequence, a preset extraction interval is used. Once the most recent image extraction is completed, the most recent extraction time is recorded as... Then the next extraction time When the time arrives At that time, the video storage server interface is called, and the video surveillance device IDs are passed in sequentially according to their numbers. The corresponding video images are then extracted, and the process is reset after extraction. and To achieve based on extraction interval The extraction process is performed at regular intervals. Furthermore, to reduce video network bandwidth consumption and considering the timeliness of video reporting, this embodiment preferably uses an extraction interval... The value ranges from 3 to 5 minutes.
[0031] like Figure 2 As shown, in the manual extraction mode when a catenary fault occurs, the preset manual extraction time for video surveillance equipment is specified. When a contact network tripping fault occurs, the operator manually selects the faulty line, substation, and power supply arm. The fault distance parameters in the tripping message from the fault locator are analyzed, and the tripping kilometer marker and the equipment location kilometer marker are spatially matched and calculated to determine the straight-line distance between the equipment and the fault point. Then, the tripping kilometer marker and tripping time calculated by the fault locator are input. Based on the logical relationship of the basic data of the five types of objects—line, section station, track, substation, and power supply arm—and the multi-dimensional data relationship pre-configured by the video surveillance equipment, a list of video surveillance equipment associated with the contact network tripping fault is obtained.
[0032] In this embodiment, based on the video surveillance device association list, video surveillance devices within a 2-kilometer range are filtered according to the trip kilometer marker and the location kilometer marker in the basic information of the video surveillance devices. The devices are then rearranged in ascending order of distance from the trip kilometer marker to form an extraction list. The start time for video image extraction is then determined to be 3 minutes before the trip time. The video storage server interface is then called, and the video surveillance device ID information and extraction start time are passed sequentially according to the order of the video surveillance devices in the extraction list. The corresponding extraction duration is then set. The video image.
[0033] In step S3, the extracted video images are automatically pushed to the video image anomaly recognition algorithm interface based on the HTTP protocol. After receiving the pushed video images, the algorithm interface calls the corresponding algorithms to perform video image analysis according to the configured sub-algorithm recognition points. The sub-algorithm recognition points include bird nests, floating objects, and tree encroachment. The video image analysis includes image segmentation, image preprocessing, image restoration, image compression, image segmentation, and image recognition and analysis.
[0034] This step can be performed based on the video analytics engine in the system, which can utilize a microservice architecture to deploy the identification items of each sub-algorithm in a containerized manner. For the bird nest detection identification item, an improved YOLO model can be used for target detection, and a dataset containing multiple bird nest samples from the catenary can be used for transfer learning during training; the floating object detection identification item can combine optical flow and background subtraction methods to identify floating foreign objects by calculating motion vectors between consecutive frames; the tree encroachment identification item uses the U-Net semantic segmentation network to perform pixel-level analysis of the vegetation coverage area around the catenary.
[0035] In this embodiment, the specific operations for video image analysis are as follows: Image segmentation: dividing the video images captured by the camera into several single-frame image data; Image preprocessing: Further reduce image noise, improve image contrast, and highlight the outlines of objects in the image after cropping; Image restoration: Filtering techniques are used to eliminate environmental noise, motion blur, overexposure, or underexposure; Image compression: Since videos contain a large number of images, image compression technology can improve image transmission speed and shorten processing time; Image segmentation: Segmenting the sub-regions that make up an image is a preliminary stage of image recognition and analysis; Image recognition and analysis: Combining the different image recognition algorithms mentioned above, as well as statistical methods, neural network methods, geometric transformation methods, etc., the obtained images are used to extract and classify features.
[0036] A sample overview of the video image analysis process could be as follows: In the image preprocessing stage, the system first performs gamma correction on the video frames, with a gamma value ranging from 1.2 to 1.5 to improve uneven lighting. Then, a non-local mean denoising algorithm is used, with a search window of 21×21 and a similarity window of 7×7 to eliminate image noise. Finally, histogram equalization is used to enhance edge features. For motion blur, Wiener filtering can be used for image restoration, with its point spread function automatically estimated based on the camera's exposure time.
[0037] In step S4, the video image anomaly recognition algorithm outputs the results of the video image analysis and returns them to the user. The output of the suspected contact network faults includes the fault image, fault type, fault occurrence time, and corresponding fault video frame number.
[0038] In step S5, the video image anomaly recognition algorithm presents suspected fault information of the overhead contact line through its output. The user performs manual analysis and confirmation based on the suspected fault information of the overhead contact line. Data that is confirmed as a real fault by manual analysis is marked as valid. After supplementing the detailed description of the real fault and the handling response information, it is stored in the overhead contact line fault database for summary. Data that is confirmed as not having a fault by manual analysis is marked as invalid and discarded.
[0039] Steps S4 and S5 together constitute the verification mechanism for the fault confirmation stage. The preferred method for this embodiment is described below: The verification mechanism can be divided into three levels. The initial verification involves the system algorithm automatically labeling the timestamps and spatial coordinates of suspected fault frames. The intermediate verification uses 3D point cloud reconstruction technology to map the 2D image onto the line BIM model to verify spatial rationality. Finally, manual verification confirms the authenticity of the fault. Confirmed valid fault data will be stored in the overhead contact line fault database. This database can use ontology modeling methods to construct fault feature vectors, thus including multi-dimensional feature fields such as fault type, meteorological conditions, and equipment model, providing data support for subsequent fault prediction and analysis. For cases where manual analysis confirms no fault has occurred, an abnormal data feedback mechanism can be used to input the relevant sample data from the manually rejected algorithm into the algorithm's retraining queue, continuously optimizing the video image anomaly recognition algorithm through incremental learning.
[0040] In a typical application scenario, an example of the method in this embodiment is as follows: When a substation detects a ground fault in phase C of a power supply arm that has lost power, the system first analyzes the trip kilometer marker K123+450 provided by the fault locator. Through the topology network, it retrieves the three tracks associated with the power supply arm and the corresponding six cameras. After calculating the spatial distance based on the camera location kilometer markers, it selects the two closest devices, K123+500 and K123+300, for video extraction. The analysis engine analyzes and identifies the video images from the three minutes prior to the fault occurrence, identifying a plastic bag wrapped around the contact wire suspension point at K123+480. After manual verification, a fault work order is generated and pushed to the maintenance system, where manual inspection immediately initiates a maintenance response. The entire process, from fault occurrence to location completion, is significantly shortened, greatly improving fault handling efficiency compared to the traditional method relying solely on manual inspection.
Claims
1. A method for identifying overhead contact line faults based on video surveillance, characterized in that, The method includes the following steps: S1. Initialize the parameters of the multidimensional database, which includes information on video surveillance equipment, video image anomaly recognition algorithms, and railway line information. S2. After parameter initialization is complete, extract the video images already collected by the specified video surveillance equipment using either timed extraction or manual extraction when a contact network fault occurs. S3. Push the extracted video images to the video image anomaly recognition algorithm that has completed parameter initialization for video image analysis; S4. After completing the video image analysis, the video image anomaly recognition algorithm outputs images of suspected contact network faults and corresponding fault type results. S5. Based on the output of the video image anomaly recognition algorithm, suspected faults in the overhead contact line are confirmed manually.
2. The contact wire fault identification method according to claim 1, characterized in that: In step S1, the parameters of the railway line information are first initialized, and then the information of the video monitoring equipment and the information of the video image anomaly recognition algorithm are configured, thereby completing the parameter initialization process of the multidimensional database as a whole. The railway line information includes the line, the section station, the track, the substation and the power supply arm. After collecting the basic data of these five types of objects, the data is entered item by item according to the logical relationship between the basic data of each type.
3. The contact wire fault identification method according to claim 2, characterized in that: The logical relationships of the basic data of the five types of objects—line, section station, track, substation, and power outage supply arm—are as follows: track → section station → line, that is, track is associated with section station, and section station is associated with line; substation → line, that is, substation is associated with line; track ← power outage supply arm → substation, that is, power outage supply arm is associated with both substation and track.
4. The contact wire fault identification method according to claim 2, characterized in that: The video surveillance equipment uses cameras and configures the basic information and multi-dimensional data relationships of the cameras. The basic information includes the camera name, camera ID, service IP, port number, and location kilometer marker. The multi-dimensional data relationships include the lines, stations, and tracks associated with the camera. The video image anomaly recognition algorithm includes multiple sub-algorithm recognition items, and each sub-algorithm recognition item is configured with a corresponding algorithm name and algorithm number.
5. The contact wire fault identification method according to claim 1, characterized in that: In step S2, the timed extraction method is as follows: for multiple video surveillance devices arranged in sequence, a preset extraction interval is used. Once the most recent image extraction is completed, the most recent extraction time is recorded as... Then the next extraction time When the time arrives At that time, the video storage server interface is called, and the video surveillance device IDs are passed in sequentially according to their numbers. The corresponding video images are then extracted, and the process is reset after extraction. and To achieve based on extraction interval The timed extraction process is performed.
6. The overhead contact line fault identification method according to claim 3, characterized in that: In step S2, the manual extraction method when a catenary fault occurs is as follows: a preset manual extraction time is set for the video surveillance equipment. When a contact network tripping fault occurs, the operator manually selects the faulty line, substation, and power outage arm, and inputs the tripping kilometer marker and tripping time calculated by the fault distance measuring instrument. Based on the logical relationship of the basic data of the five types of objects—line, section station, track, substation, and power outage arm—and the multi-dimensional data relationship pre-configured by the video monitoring equipment, a list of associated video monitoring equipment related to the contact network tripping fault is obtained.
7. The overhead contact line fault identification method according to claim 6, characterized in that: Based on the video surveillance device association list, according to the trip kilometer marker and the location kilometer marker in the basic information of the video surveillance devices, video surveillance devices that meet the preset distance threshold are filtered. These devices are then re-arranged in ascending order of distance from the trip kilometer marker to form an extraction list. The start time for video image extraction is then set to a preset duration before the trip time. The video storage server interface is then called, and the video surveillance device IDs and extraction start times are passed in sequentially according to the order of the video surveillance devices in the extraction list, with the corresponding extraction duration set. The video image.
8. The contact wire fault identification method according to claim 1, characterized in that: In step S3, the extracted video images are automatically pushed to the video image anomaly recognition algorithm interface based on the HTTP protocol. After receiving the pushed video images, the algorithm interface calls the corresponding algorithms to perform video image analysis according to the configured recognition points of each sub-algorithm. The video image analysis includes image segmentation, image preprocessing, image restoration, image compression, image segmentation, image recognition and analysis.
9. The overhead contact line fault identification method according to claim 1, characterized in that: In step S4, the video image anomaly recognition algorithm outputs the results of the video image analysis and returns them to the user. The output of the suspected contact network faults includes the fault image, fault type, fault occurrence time, and corresponding fault video frame number.
10. The overhead contact line fault identification method according to claim 1, characterized in that: In step S5, the video image anomaly recognition algorithm presents suspected contact network fault information through its output, and the user performs manual analysis and confirmation based on the suspected contact network fault information. Data that has been manually analyzed and confirmed as genuine faults are marked as valid. After supplementing the detailed description of the genuine fault and the handling response information, the data is stored in the overhead contact line fault database for aggregation. Data that was confirmed by manual analysis to be without faults was marked as invalid and discarded.