Train pantograph detection system and method based on artificial intelligence

Abstract

The invention provides a train pantograph detection system and method based on artificial intelligence, and the system comprises a plurality of camera devices which are installed in a train detection room in a dispersed manner, and are used for collecting the image data of a train identification part and a pantograph part on a train entering the train detection room; the edge analysis device is connected with the camera device and is used for processing the image data based on a pre-trained image analysis model, identifying a train identifier, obtaining a pantograph analysis result and sending the pantograph analysis result to the cloud platform; and the cloud platform is connected with the edge analysis device, and is used for storing the pantograph analysis results in a classified manner according to the train identifiers, and generating alarm information when the pantograph analysis results have pantograph abnormities. The detection efficiency is improved, the subjectivity of manual detection is reduced, the accuracy and reliability of a detection result are improved, the standardization of a detection process is realized, and the situations of manual misjudgment, missing detection and the like are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of train pantograph and fault detection, and in particular to a train pantograph detection system and method based on artificial intelligence. Background Art

[0002] As a key component to ensure power supply and locomotive operation safety, the accuracy and efficiency of locomotive pantograph detection are crucial to the safety of the entire railway system. However, traditional manual detection methods have the following problems:

[0003] (1) There are many manual inspection links, which cannot meet the needs of large-scale inspection. It is inefficient, time-consuming and labor-intensive. In addition, manual inspection is labor-intensive and easily leads to missed inspections and misjudgments.

[0004] (2) The accuracy of the test is affected by the experience and working status of the test personnel. The test results are biased towards the subjective judgment of the personnel, and the consistency and reliability are difficult to guarantee;

[0005] (3) Manual inspections often have incomplete records, making data recording and historical tracing difficult;

[0006] (4) There is no uniform testing standard for each employee, and there is a lack of scientific and standardized testing methods. Summary of the invention

[0007] Based on the above problems, the present invention provides a train pantograph detection system and method based on artificial intelligence, aiming to solve the technical problems in the prior art that manual pantograph detection is prone to omission, misjudgment and wrong judgment.

[0008] A train pantograph detection system based on artificial intelligence, comprising:

[0009] A plurality of camera devices are dispersedly installed in the train inspection room, and are used to collect image data of the train identification part and the pantograph part of the train entering the train inspection room;

[0010] The edge analysis device is connected to the camera device and is used to process the image data based on the pre-trained image analysis model, identify the train logo, obtain the pantograph analysis results and send them to the cloud platform;

[0011] Cloud platform, connected to edge analytics devices, for:

[0012] Store pantograph analysis results by train identification;

[0013] When pantograph abnormality is found in the pantograph analysis results, an alarm message is generated.

[0014] Further, the edge analysis device includes:

[0015] A storage server, used for storing image data collected by the camera device;

[0016] The edge analysis device is connected to the storage server and is used to process the image data stored in the storage server based on the pre-trained image analysis model, identify the train logo, obtain the pantograph analysis results and send them to the cloud platform.

[0017] Furthermore, the edge analysis device includes:

[0018] The target detection module is used to identify the target object in the image data based on the target recognition model and segment the image of the target object;

[0019] The abnormality analysis module is connected to the target detection module and is used to: perform abnormality analysis on the target object based on the abnormality analysis model corresponding to the target object, and when the target object is analyzed to be abnormal, an abnormality is marked in the image of the target object to obtain the analysis result of the target object;

[0020] The result sending module is connected to the abnormal analysis module and is used to integrate the analysis results of each target object on the pantograph into a pantograph analysis result and send it to the cloud platform.

[0021] Furthermore, the cloud platform includes:

[0022] The cloud storage module stores the analysis results of each target object of the pantograph according to the train identification classification and time sequence;

[0023] An alarm generation module, connected to the cloud storage module, is used to generate alarm information for the analysis result of the target object with abnormality, and the alarm information includes an abnormal annotated image of the target object with abnormality;

[0024] The display module is connected to the alarm generation module and is used to display the alarm information.

[0025] Furthermore, the cloud platform includes:

[0026] The cloud storage module stores the analysis results of each target object of the pantograph according to the train identification classification and time sequence, forming a time series of the analysis results of each target object;

[0027] A risk prediction module, connected to the cloud storage module, is used to use a pre-trained risk prediction model to perform risk prediction on the time series of analysis results of each target object, obtain the abnormal risk probability of each target object, and obtain the abnormal risk probability of the pantograph based on the abnormal risk probability of each target object;

[0028] The display module is connected to the risk prediction module and is used for the abnormal risk probability of each target object and the abnormal risk probability of the pantograph.

[0029] A train pantograph detection method based on artificial intelligence, using the aforementioned train pantograph detection system based on artificial intelligence, comprising:

[0030] Step A1, a camera device collects image data of a train identification part and a pantograph part on a train entering a train inspection room;

[0031] Step A2: The edge analysis device processes the image data based on the pre-trained image analysis model, identifies the train logo, obtains the pantograph analysis results, and sends them to the cloud platform;

[0032] Step A3, the cloud platform stores the pantograph analysis results according to the train identification classification, and generates an alarm message when there is a pantograph abnormality in the pantograph analysis results.

[0033] Further, the edge analysis apparatus includes a storage server and an edge analysis device, and step A2 includes:

[0034] Step A21, the storage server stores the image data collected by the camera device;

[0035] Step A22, the edge analysis device processes the image data stored in the storage server based on the pre-trained image analysis model, identifies the train logo, obtains the pantograph analysis results and sends them to the cloud platform.

[0036] Further, step A22 includes:

[0037] Step A221, identifying a target object in the image data based on a target recognition model and segmenting an image of the target object;

[0038] Step A222, performing an abnormality analysis on the target object based on the abnormality analysis model corresponding to the target object, and when the target object is analyzed to be abnormal, marking the abnormality in the image of the target object to obtain an analysis result of the target object;

[0039] Step A223, integrating the analysis results of each target object on the pantograph into a pantograph analysis result and sending it to the cloud platform.

[0040] Further, step A3 includes:

[0041] Step A31, storing the analysis results of each target object of the pantograph according to the train identification classification and time sequence;

[0042] Step A32, generating alarm information for the analysis result of the target object with abnormality, the alarm information including the abnormality annotated image of the target object with abnormality;

[0043] Step A33, display the alarm information.

[0044] Further, in step A31, the analysis results of each target object of the pantograph are stored according to the train identification classification and time sequence to form a time series of the analysis results of each target object;

[0045] After step A31, the following steps are also included:

[0046] Step B1, using a pre-trained risk prediction model to perform risk prediction on the analysis result time series of each target object, deriving the abnormal risk probability of each target object, and deriving the abnormal risk probability of the pantograph based on the abnormal risk probability of each target object;

[0047] In step A33, the abnormal risk probability of each target object and the abnormal risk probability of the pantograph are displayed.

[0048] The beneficial technical effects of the present invention are as follows: pantograph fault detection is performed based on an artificial intelligence algorithm to replace manual detection, thereby improving detection efficiency, reducing the subjectivity of manual detection, improving the accuracy and reliability of detection results, realizing standardization of the detection process, and reducing human misjudgment and missed detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1-3 A schematic diagram of a module of a train pantograph detection system based on artificial intelligence according to the present invention;

[0050] Figure 4-8 A flow chart of the steps of a train pantograph detection method based on artificial intelligence of the present invention;

[0051] Fig. 9 This is an installation position diagram of a camera device of an artificial intelligence-based train pantograph detection system of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0054] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0055] See also Figure 1 The present invention provides a train pantograph detection system based on artificial intelligence, comprising:

[0056] A plurality of camera devices (1) are dispersedly installed in the train inspection room and are used to collect image data of train identification parts and pantograph parts on the train entering the train inspection room;

[0057] An edge analysis device (2) is connected to the camera device (1) and is used to process image data based on a pre-trained image analysis model, identify train identification, obtain pantograph analysis results, and send them to a cloud platform;

[0058] The cloud platform (3), connected to the edge analysis device (2), is used to:

[0059] Store pantograph analysis results by train identification;

[0060] When pantograph abnormality is found in the pantograph analysis results, an alarm message is generated.

[0061] The implementation of the present invention will significantly improve the efficiency and accuracy of train pantograph detection. When a train passes through a train inspection shed (train inspection room) at a speed of 3-5km per hour, the train number and pantograph can be effectively and accurately identified, meeting the identification requirements of the pantograph dynamic detection system for train number and pantograph.

[0062] Through automated image acquisition and advanced image recognition technology, the detection time is greatly shortened, thereby improving production efficiency. The deep learning algorithm adopted by the system reduces the interference of human factors and improves the reliability of the detection results. In addition, the unified detection standards and data-driven decision-making process realize the standardization of the detection process and optimize the maintenance process. Through real-time data recording and historical tracing functions, data management capabilities are enhanced. The automation of the system reduces labor costs and reduces economic losses caused by detection errors, providing a strong guarantee for the safety and reliability of railway transportation.

[0063] See also Figure 2 , further, the edge analysis device includes:

[0064] A storage server (21), used for storing image data collected by the camera device;

[0065] The edge analysis device (22) is connected to the storage server (21) and is used to process the image data stored in the storage server based on a pre-trained image analysis model, identify the train logo, obtain the pantograph analysis results and send them to the cloud platform.

[0066] This system adopts a layered design and consists of three main parts: data collection layer, edge analysis layer and cloud service layer, ensuring the efficiency and orderliness of the detection process.

[0067] The data collection layer has multiple cameras installed in the train inspection room, which are responsible for capturing the image data of the train number (train identification) and its pantograph in real time, and transmitting this data to the edge analysis layer in real time.

[0068] The camera device is, for example, a network high-definition camera, which may be an infrared camera or a visible light camera.

[0069] There are multiple camera devices deployed in the train inspection room. They can be deployed on the front, middle and rear sections of the train inspection room respectively. The deployment positions of the camera devices in these three sections are the same.

[0070] See also Fig. 9 Specifically, on each section, there are two cameras on the inclined surface of the ceiling in the train inspection room, one on the left wall and one on the right wall, for a total of four. The two cameras on the ceiling monitor the upper part of the pantograph, and the cameras on the left and right walls monitor the sides and bottom of the pantograph. It can be seen that four cameras are deployed in each section of the front, middle and three sections, and a total of 12 cameras are deployed in the train inspection room.

[0071] In order to ensure that the camera can clearly capture the moving train logo (i.e. train number) and pantograph components, there needs to be sufficient light source in the train inspection room. Therefore, when deploying the camera, it is necessary to set up a supplementary light source in the inspection room. The light source is installed next to each camera to ensure that the illumination at the top of the train in the train inspection room can reach more than 300lx, especially in insufficient light at night, to ensure that each camera can generate a clear image.

[0072] The camera device at each position needs to be adjusted in angle. By adjusting the focal length, angle, etc., the camera devices at different positions can have clear detection focus. For example, the top camera device takes pictures of the car number and the charcoal slide plate, and the side camera device takes pictures of the screws at each pantograph part, etc. The camera device shooting imaging requires that the shooting target is clear and the focus is clear when the train is moving. The target position can be obtained from the continuous shooting video stream. High-definition pictures with high picture quality can be obtained.

[0073] Edge analysis layer: This layer consists of visual AI edge analysis devices and storage servers between devices. The visual AI edge analysis devices are responsible for deep calculation and analysis of images from the data acquisition layer, and then pass the analysis results to the cloud service layer.

[0074] Cloud service layer: A cloud platform for pantograph damage quality inspection, assessment, monitoring and early warning is deployed in the control room. The cloud platform is responsible for receiving the analysis results sent by the edge analysis layer, and performing key functions such as abnormal alarm, notification of responsible personnel, risk prediction, model update, distribution strategy, false alarm library maintenance and configuration management of edge devices.

[0075] See also Figure 2 , further, the edge analysis device (22) comprises:

[0076] A target detection module (221), used to identify a target object in the image data based on a target recognition model and segment an image of the target object;

[0077] An abnormality analysis module (222) is connected to the target detection module (221) and is used to: perform abnormality analysis on the target object based on an abnormality analysis model corresponding to the target object, and when an abnormality is found in the target object, an abnormality is marked in the image of the target object to obtain an analysis result of the target object;

[0078] The result sending module (223) is connected to the abnormality analysis module (222) and is used to integrate the analysis results of each target object on the pantograph into a pantograph analysis result and send it to the cloud platform.

[0079] The edge analysis device (22) uses deep learning technology to identify specific objects in the image and determine their locations, and classifies the detected specific objects to identify which predefined category they belong to. For example, it is a carbon slide, a bolt area, or a wire rope area. Anomaly analysis is performed based on the specific type of the identification book. For example, if the category is a carbon slide, analyze whether the carbon slide has cracks, breakage, and other anomalies, such as bolt area, whether there are bolts falling off, and whether there are wire rope strands in the wire rope area.

[0080] Specifically, the target object is a target detection object of the pantograph, such as a carbon slide plate, a steel wire rope, and a bolt fastening area, etc. The target detection module (221) is also used to identify a train logo.

[0081] Cloud platforms are also used to:

[0082] Acquire several pantograph images as training samples;

[0083] Initial construction of image analysis model;

[0084] The constructed image analysis model is trained using training samples to preliminarily establish an image analysis model.

[0085] The image analysis model of the present invention (including the target recognition model and the abnormal analysis model) is combined with the basic models such as image classification, object detection, image comparison, and image segmentation for the first construction. After that, a large number of positive and negative training samples are collected, and the characteristics of abnormal phenomena occurring in the monitoring target object are obtained from these samples using feature extraction technology to train the model. The purpose of training is to perform secondary development and optimization of the model algorithm based on the characteristics of the monitoring target object, perform a large number of operations and training of the model through the training sample set, and continuously adjust the parameters to improve the recognition accuracy. The main parameters adjusted include the estimator, the minimum number of sample segmentations, the learning rate, the loss amount, the maximum depth, the number of iterations, the feature score, the threshold, the type, the range, the weight and the bias, etc. Through continuous iterative operations, until the model reaches the allowed accuracy and precision within the training sample set.

[0086] After the model is initially established, these test sample sets are used to test the accuracy and precision of the visual AI security detection model. Once deviations are found, the model algorithm is optimized and the parameters are adjusted until the optimal performance is achieved within the test sample set.

[0087] In the pre-launch stage, the trained image analysis model is loaded onto the edge analysis device for online testing. At the same time, a false alarm library is established to collect samples of false alarms, missed alarms, and deviations. The model is adjusted and optimized on the cloud platform to solve these problems. The adjusted and optimized model is upgraded and updated online, and loaded onto the edge analysis device again for online operation testing. Through continuous iteration and repetition, the model can be continuously optimized according to the actual production situation.

[0088] See also Figure 3 , further, the cloud platform (3) includes:

[0089] A cloud storage module (31) stores the analysis results of each target object of the pantograph according to the train identification classification and time sequence;

[0090] An alarm generation module (32) connected to the cloud storage module (31) is used to generate alarm information for the analysis result of the target object with abnormality, wherein the alarm information includes an abnormal annotated image of the target object with abnormality;

[0091] The display module (33) is connected to the alarm generation module (32) and is used to display the alarm information.

[0092] The display module (33) is also connected to the cloud storage module (31), and displays not only the alarm information but also the analysis results of each target object of the pantograph, wherein the analysis results of each target object include an image of each target object. If the analysis result is abnormal, the displayed image is an image of the target object with abnormality marked. The abnormality marking can be to use a frame to outline the area where the abnormality occurs, for example, when the carbon slide plate is cracked or damaged, the crack area or the damaged area is marked, and when the bolt is detached, the area where the bolt is detached is marked.

[0093] As an embodiment of anomaly detection of the present invention, for example, the target detection module (221) of the edge analysis device preliminarily classifies the images collected by the camera device and obtains the target object of the carbon slide, then enters the carbon slide damage detection branch, the abnormal analysis module (222) locates the carbon slide, and uses the corresponding abnormal analysis model (a target detection algorithm) on the carbon slide component to identify the damage. If the detection locates a damaged part on the carbon slide, the damaged position is marked, and the cloud platform uses the damaged mark image as an alarm image to form an alarm information and push it to relevant personnel and the display module. The carbon slide is a component that is easily damaged in the pantograph component, and is prone to cracks and damage. It is easy to clearly capture it from the top camera device.

[0094] As another embodiment of abnormality detection of the present invention, for example, the target detection module (221) of the edge analysis device preliminarily classifies the image collected by the camera device to obtain the target object of the bolt fastening area of ​​the pantograph, and then enters the bolt fastening area detection branch. The abnormality analysis module (222) locates the position of the bolt fastening area, first detects the bolts and nuts located at the bolt fastening area of ​​the pantograph with the abnormality analysis model (a machine vision image detection algorithm), and then compares and judges according to the preset position information to determine whether it meets the preset number and position of the bolt fastening area. If the number of bolts does not match, it is determined to be an abnormality such as bolt falling off or missing. Depending on the connection parts, there can be multiple bolt fastening areas.

[0095] The bolts and nuts of the pantograph are the key points for connecting various components. There are many of them, and the targets are small, so they are easily blocked and unclear in the photos. The shedding and loosening of screws are difficult to detect in the entire pantograph detection of running trains, and it is difficult to fully cover the scenes. Based on this actual situation, it is first necessary to ensure that the bolts and nuts of the key parts can be clearly captured by the camera installation position, and the key positions are arranged. Instead of pursuing wide-angle shooting of the bolt and nut positions, the key positions are not detected, the captured images are not good, and the detection requirements are not met, which is not worth the loss.

[0096] Furthermore, the cloud platform (3) includes:

[0097] A cloud storage module (31) stores the analysis results of each target object of the pantograph according to the train identification classification and time sequence, forming a time series of the analysis results of each target object;

[0098] A risk prediction module (34), connected to the cloud storage module (31), is used to use a pre-trained risk prediction model to perform risk prediction on the time series of analysis results of each target object, obtain the abnormal risk probability of each target object, and obtain the abnormal risk probability of the pantograph based on the abnormal risk probability of each target object;

[0099] The display module (33) is connected to the risk prediction module (34) and is used for the abnormal risk probability of each target object and the abnormal risk probability of the pantograph.

[0100] The system's preventive maintenance and risk management functions help to promptly detect and resolve potential problems, ensuring safe train operation.

[0101] The cloud platform of the present invention is mainly used to generate alarm information on the one hand and to perform risk analysis and prediction on the other hand. It is also used to build and train, update and maintain image analysis models, as well as to configure and manage edge devices.

[0102] See also Figure 4 The present invention also provides a train pantograph detection method based on artificial intelligence, using the aforementioned train pantograph detection system based on artificial intelligence, comprising:

[0103] Step A1, a camera device collects image data of a train identification part and a pantograph part on a train entering a train inspection room;

[0104] Step A2: The edge analysis device processes the image data based on the pre-trained image analysis model, identifies the train logo, obtains the pantograph analysis results, and sends them to the cloud platform;

[0105] Step A3, the cloud platform stores the pantograph analysis results according to the train identification classification, and generates an alarm message when there is a pantograph abnormality in the pantograph analysis results.

[0106] Through automated image acquisition and advanced image recognition technology, the detection time is greatly shortened, thereby improving production efficiency. The deep learning algorithm adopted by the system reduces the interference of human factors and improves the reliability of the detection results. In addition, the unified detection standards and data-driven decision-making process realize the standardization of the detection process and optimize the maintenance process. Through real-time data recording and historical tracing functions, data management capabilities are enhanced. The automation of the system reduces labor costs and reduces economic losses caused by detection errors, providing a strong guarantee for the safety and reliability of railway transportation.

[0107] Specifically, the camera device is, for example, a network high-definition camera, which may be an infrared camera or a visible light camera.

[0108] Specifically, there are multiple camera devices deployed in the train inspection room, and can be deployed on the front, middle and rear sections of the train inspection room respectively. The deployment positions of the camera devices in these three sections are the same.

[0109] Specifically, on each cross section, there are four cameras deployed on two inclined surfaces of the ceiling in the train inspection room, one on the left wall, and one on the right wall.

[0110] Specifically, a light source is installed next to each camera.

[0111] See also Figure 5 , further, the edge analysis device includes a storage server and an edge analysis device, and step A2 includes:

[0112] Step A21, the storage server stores the image data collected by the camera device;

[0113] Step A22, the edge analysis device processes the image data stored in the storage server based on the pre-trained image analysis model, identifies the train logo, obtains the pantograph analysis results and sends them to the cloud platform.

[0114] See also Figure 6 , further, step A22 includes:

[0115] Step A221, identifying a target object in the image data based on a target recognition model and segmenting an image of the target object;

[0116] Step A222, performing an abnormality analysis on the target object based on the abnormality analysis model corresponding to the target object, and when the target object is analyzed to be abnormal, marking the abnormality in the image of the target object to obtain an analysis result of the target object;

[0117] Step A223, integrating the analysis results of each target object on the pantograph into a pantograph analysis result and sending it to the cloud platform.

[0118] Specifically, the target object is a target detection object of the pantograph, such as a carbon slide plate, a steel wire rope, a bolt fastening area, and the like.

[0119] See also Figure 7 , further, step A3 comprises:

[0120] Step A31, storing the analysis results of each target object of the pantograph according to the train identification classification and time sequence;

[0121] Step A32, generating alarm information for the analysis result of the target object with abnormality, the alarm information including the abnormality annotated image of the target object with abnormality;

[0122] Step A33, display the alarm information.

[0123] Specifically, in step A33, not only the alarm information is displayed, but also the analysis results of each target object of the pantograph are displayed. The analysis results of each target object include the image of each target object. If the analysis result is abnormal, the displayed image is the image of the target object with abnormality marked. The abnormality marking can be to use a frame to outline the area where the abnormality occurs, for example, when the carbon slide plate is cracked or damaged, the crack area or damaged area is marked, and when the bolt falls off, the area where the bolt falls off is marked.

[0124] See also Figure 8 , further, in step A31, the analysis results of each target object of the pantograph are stored according to the train identification classification and time sequence to form a time series of the analysis results of each target object;

[0125] After step A31, the following steps are also included:

[0126] Step B1, using a pre-trained risk prediction model to perform risk prediction on the analysis result time series of each target object, deriving the abnormal risk probability of each target object, and deriving the abnormal risk probability of the pantograph based on the abnormal risk probability of each target object;

[0127] In step A33, the abnormal risk probability of each target object and the abnormal risk probability of the pantograph are displayed.

[0128] The system's preventive maintenance and risk management functions help to promptly detect and resolve potential problems, ensuring safe train operation.

[0129] The introduction of image recognition technology provides an efficient and accurate solution for intelligent detection of locomotive pantographs. Through automated image acquisition and advanced image processing algorithms, the intelligent detection system can quickly identify and analyze pantograph defects, thereby greatly improving detection efficiency, reducing human errors, and ensuring the consistency and traceability of detection results. This not only helps to improve the safety and reliability of railway transportation, but also provides important technical support for railway maintenance and operation, and promotes technological progress and modernization in the railway industry.

[0130] The above are only preferred embodiments of the present invention, and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A train pantograph detection system based on artificial intelligence, characterized in that: include: A plurality of camera devices are dispersedly installed in the train inspection room, and are used to collect image data of train identification parts and pantograph parts on the train entering the train inspection room; An edge analysis device, connected to the camera device, for processing the image data based on a pre-trained image analysis model, identifying the train logo and obtaining the pantograph analysis result and sending it to the cloud platform; A cloud platform, connected to the edge analysis device, is used to: Classify and store the pantograph analysis results according to the train identification; When pantograph abnormality is found in the pantograph analysis result, an alarm message is generated.

2. The train pantograph detection system based on artificial intelligence as claimed in claim 1, characterized in that: The edge analysis device comprises: A storage server, used for storing image data collected by the camera device; The edge analysis device is connected to the storage server and is used to process the image data stored in the storage server based on a pre-trained image analysis model, identify the train logo, obtain the pantograph analysis results and send them to the cloud platform.

3. The train pantograph detection system based on artificial intelligence as claimed in claim 2, characterized in that: The edge analysis device comprises: A target detection module, used to identify a target object in the image data and segment an image of the target object based on a target recognition model; an abnormality analysis module, connected to the target detection module, for: performing abnormality analysis on the target object based on an abnormality analysis model corresponding to the target object, and when an abnormality is found in the target object, marking the abnormality in the image of the target object to obtain an analysis result of the target object; A result sending module is connected to the abnormality analysis module and is used to integrate the analysis results of each target object on the pantograph into a pantograph analysis result and send it to the cloud platform.

4. The train pantograph detection system based on artificial intelligence as claimed in claim 3, characterized in that: The cloud platform includes: A cloud storage module stores the analysis results of each target object of the pantograph according to the train identification classification and time sequence; An alarm generation module, connected to the cloud storage module, for generating the alarm information for the analysis result of the target object with abnormality, wherein the alarm information includes an abnormal annotated image of the target object with abnormality; A display module is connected to the alarm generation module and is used to display the alarm information.

5. The train pantograph detection system based on artificial intelligence as claimed in claim 3, characterized in that: The cloud platform includes: A cloud storage module stores the analysis results of each target object of the pantograph according to the train identification classification and time sequence to form a time series of the analysis results of each target object; A risk prediction module, connected to the cloud storage module, for using a pre-trained risk prediction model to perform risk prediction on the time series of analysis results of each target object, deriving an abnormal risk probability of each target object, and deriving an abnormal risk probability of the pantograph based on the abnormal risk probability of each target object; A display module is connected to the risk prediction module and is used for the abnormal risk probability of each target object and the abnormal risk probability of the pantograph.

6. A train pantograph detection method based on artificial intelligence, characterized in that: A train pantograph detection system based on artificial intelligence as described in any one of claims 1 to 5, comprising: Step A1, a camera device collects image data of a train identification part and a pantograph part on a train entering the train inspection room; Step A2: the edge analysis device processes the image data based on a pre-trained image analysis model, identifies the train logo, obtains the pantograph analysis results, and sends them to the cloud platform; Step A3, the cloud platform stores the pantograph analysis results according to the train identification classification, and generates an alarm message when there is a pantograph abnormality in the pantograph analysis results.

7. The train pantograph detection method based on artificial intelligence as claimed in claim 6, characterized in that: The edge analysis device includes a storage server and an edge analysis device, and the step A2 includes: Step A21, the storage server stores the image data collected by the camera device; Step A22, the edge analysis device processes the image data stored in the storage server based on a pre-trained image analysis model, identifies the train logo, obtains the pantograph analysis results and sends them to the cloud platform.

8. The train pantograph detection method based on artificial intelligence as claimed in claim 7, characterized in that: The step A22 comprises: Step A221, identifying a target object in the image data based on a target recognition model and segmenting an image of the target object; Step A222, performing an abnormality analysis on the target object based on an abnormality analysis model corresponding to the target object, and when an abnormality is found in the target object, an abnormality is marked in the image of the target object to obtain an analysis result of the target object; Step A223, integrating the analysis results of each target object on the pantograph into a pantograph analysis result and sending it to the cloud platform.

9. The train pantograph detection method based on artificial intelligence as claimed in claim 8, characterized in that: The step A3 comprises: Step A31, storing the analysis results of each target object of the pantograph according to the train identification classification and time sequence; Step A32, generating the alarm information for the analysis result of the target object with abnormality, wherein the alarm information includes an abnormal annotated image of the target object with abnormality; Step A33, display the alarm information.

10. The train pantograph detection method based on artificial intelligence as claimed in claim 9, characterized in that: In the step A31, the analysis results of each target object of the pantograph are stored according to the train identification classification and time sequence to form a time series of the analysis results of each target object; After step A31, the following steps are also included: Step B1, using a pre-trained risk prediction model to perform risk prediction on the time series of the analysis results of each target object, to obtain the abnormal risk probability of each target object, and to obtain the abnormal risk probability of the pantograph based on the abnormal risk probability of each target object; In the step A33, the abnormal risk probability of each target object and the abnormal risk probability of the pantograph are displayed.

Images