A rail transit train image collection and distribution detection system

Abstract

The present invention relates to a rail transit train image collection and distribution detection system, comprising: multiple field devices and a server; each field device comprises: multiple image acquisition modules, a local industrial computer, a first communication module, and a business presentation module; the server comprises: a second communication module and a detection module; multiple image acquisition modules are respectively installed in designated areas beside designated tracks, and transmit the collected images to the local industrial computer; the local industrial computer performs a first mosaic of the images collected during the inspection to generate overall image information of the train, and sends the image information and the train's operation information to the server via the first communication module; the detection module performs disease category detection on the overall image information based on the detection strategy combined with the operation information, and sends the detection results to the business presentation module via the second communication module. The above system can centrally process the collected image data of different trains, improving the image restoration accuracy and detection accuracy of the entire vehicle.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a rail transit train image collection and distribution detection system. Background Art

[0002] In rail transit, train exterior image inspection typically involves real-time detection of faults such as paint damage, detached covers, loose cables, and loose bolts. This provides a basis for train inspection and maintenance, ensuring the safe operation of trains on a daily basis. Currently, existing technologies for train exterior image inspection utilize image processing to replace manual visual inspection. This involves capturing images of visible components as the train passes through; these images are then used to restore the entire train's appearance; and finally, the images are analyzed, compared, and used to determine faults, enabling train exterior image inspection.

[0003] However, due to the low speed measurement accuracy and frequency of the on-site trackside speed measurement system and the low accuracy of vehicle image restoration during train appearance image detection, the final detection accuracy is low; in addition, as the train ages, a local detection server with a fixed algorithm cannot adapt to the development and changes in the train appearance. Summary of the Invention

[0004] (1) Technical issues to be resolved

[0005] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a rail transit train image collection and distribution detection system, which can improve the accuracy of vehicle image restoration and improve the long-term detection accuracy of the train appearance image detection system.

[0006] (2) Technical solution

[0007] In order to achieve the above objectives, the main technical solutions adopted by the present invention include:

[0008] In a first aspect, an embodiment of the present invention provides a rail transit train image collection and distribution detection system, comprising:

[0009] multiple field devices and a server;

[0010] Each on-site device includes: multiple image acquisition modules, a local industrial computer, a first communication module and a business presentation module; the server includes: a second communication module and a detection module;

[0011] Multiple image acquisition modules are respectively installed in designated areas beside the designated track to transmit the acquired images to the local industrial computer; all image acquisition modules are electrically connected to the local industrial computer;

[0012] The local industrial computer performs a first puzzle based on all images and vehicle speed information transmitted by the connected image acquisition module to generate overall image information of the train on the current track and operation information of the train, and sends the image information and operation information to the server through the first communication module;

[0013] The second communication module receives the overall image information and operation information and transmits them to the detection module. The detection module performs disease category detection on the overall image information based on the detection strategy and the operation information to obtain the detection results; and transmits them to the business presentation module of the corresponding on-site device through the second communication module and the first communication module. The business presentation module displays the detection results.

[0014] Optionally, one image acquisition module is installed at the top of the track, one image acquisition module is installed on the side of the vehicle next to the track, and one image acquisition module is installed on the bottom of the vehicle on the track; one image acquisition module is installed in the track area to capture the train speed-time table, and one image acquisition module is installed in the track area to capture the train number;

[0015] Each image acquisition module is a high-definition high-frequency line array camera / high-definition high-speed 4K camera;

[0016] Each image acquisition module takes pictures at a fixed frame rate.

[0017] Optionally, the local industrial computer is specifically used for

[0018] Identify the feature points in each image, determine the location of the pixels in the original train image based on the pixel location and speed-time table of each feature point, and piece together all the images to form the overall image information of the train on the current track;

[0019] Identify partial information of the feature points of the image and determine the train operation information;

[0020] or,

[0021] The local industrial computer uses a deep learning network to identify feature points in each image. Based on the pixel location and speed-time table of each feature point, it determines the location of the pixel in the original train image and then pieces together all the images to create a comprehensive image of the train currently on the track.

[0022] Identify the speed-timetable and partial information of feature points in the image to determine the train operation information;

[0023] The train operation information includes one or more of the following: train number, train speed, train model, detection unique serial number code, business instruction code, and user unit code.

[0024] Optionally, the local industrial computer is specifically used for

[0025] The recognition image acquisition module collects characteristic points (train head, train tail and middle key points) in the image, receives the speed-timetable, and determines the train operation information;

[0026] According to the speed v in the train operation information, the acquisition frequency P of the image acquisition module and the width L of the acquisition frame;

[0027] Based on the frame number step of (L / v) / 2, multiple frames of images are extracted from the feature point interval image, and redundant image splicing between feature points is performed; the images are spliced ​​in a connected manner according to the image acquisition time point to generate the overall image information of the train.

[0028] Optionally, the server further includes: an expert experience module;

[0029] The expert experience module is used to regularly collect the real disease detection results and false disease detection results of the same vehicle model as the expert experience database;

[0030] The expert experience database is used to periodically correct the experience values ​​used by the detection module;

[0031] Furthermore, the expert experience database is used to provide reference information of real disease detection results and false disease detection results used by the detection module in similarity comparison detection;

[0032] The expert experience database has one expert experience database for each vehicle model.

[0033] Optionally, the detection module is specifically used to

[0034] Based on the train operation information, selecting a train template to which the overall image information belongs;

[0035] The overall image information is used as a redundant image, and the redundant image is refined by means of a train template and a superpixel segmentation method or a semantic segmentation method, so that the redundant image is adjusted to a refined image corresponding to the overall image information;

[0036] Perform template matching on the refined image and the train template to obtain a detection result; or, match the refined image to a pre-trained deep network learning model to obtain a detection result output by the deep network learning model;

[0037] The detection results are classified into diseases and suspected diseases, and the similarity between the suspected diseases and the real diseases and misdetected diseases in the expert experience database is calculated to determine whether the suspected diseases are real diseases.

[0038] Optionally, the business presentation module includes:

[0039] An alarm unit, configured to send out an alarm signal when the detection result is a detection result of a serious damage level;

[0040] An analysis unit is used to analyze and process the test results within a preset time period to obtain report information and life cycle prediction information;

[0041] The display unit is used to display each test result.

[0042] Optionally, the detection module includes:

[0043] A vehicle type screening unit, based on the train operation information, determines the train type An to which the overall image information currently sent by the on-site device belongs and the train template to which the train type An belongs;

[0044] The An model appearance detection unit performs matching based on the train template to which the train model An belongs and the overall image information to obtain a detection result.

[0045] Optionally, the field device further includes:

[0046] Speed ​​measuring unit, used to measure the running speed of the train passing through the on-site device;

[0047] A fill light source, used for providing fill light when the image acquisition module is acquiring images;

[0048] A trigger unit is used to start the field device.

[0049] Optionally, the second communication module and the first communication module both include:

[0050] Firewall unit, used to encrypt and decrypt data sent to the external network;

[0051] A transmission unit is used for transmission by means of optical communication or by means of 5G communication.

[0052] (3) Beneficial effects

[0053] The distributed detection system of the present invention is designed based on the concept of "distributed collection and centralized detection". It uses a professional detection server for centralized detection and processes the collected image data of different trains to improve the image restoration accuracy and detection accuracy of the entire vehicle.

[0054] The server of this embodiment includes a secondary correction algorithm for the whole vehicle image to improve the accuracy of the whole vehicle image restoration; it also has the ability to iterate the algorithm and summarize experience to improve the long-term detection accuracy of the train appearance image detection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 and Figure 2 They are respectively schematic diagrams of the architecture of a rail transit train image collection and distribution detection system provided by an embodiment of the present invention;

[0056] Figure 3 This is a schematic structural diagram of a field device according to an embodiment;

[0057] Figure 4 This is a schematic diagram of the structure of a business presentation module according to an embodiment;

[0058] Figure 5 A schematic diagram of the structure of a high-speed communication module provided in one embodiment;

[0059] Figure 6a A schematic diagram of an image template corresponding to a redundant image provided by an embodiment;

[0060] Figure 6b A schematic diagram of a redundant image refinement process provided by an embodiment. DETAILED DESCRIPTION

[0061] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation methods in conjunction with the accompanying drawings.

[0062] Example 1

[0063] like Figure 1 As shown, an embodiment of the present invention provides a rail transit train image collection and distribution detection system. The system of this embodiment includes: multiple field devices and a server;

[0064] Each on-site device includes: multiple image acquisition modules, a local industrial computer, a first communication module and a business presentation module; the server includes: a second communication module and a detection module;

[0065] Multiple image acquisition modules are respectively installed in designated areas beside the designated track to transmit the acquired images to the local industrial computer; all image acquisition modules are electrically connected to the local industrial computer;

[0066] The local industrial computer performs a first mosaic based on all images and train speed information collected during the inspection and transmitted by the connected image acquisition module to generate overall image information of the train and operation information of the train, and sends the overall image information and operation information to the server via the first communication module;

[0067] The second communication module receives the overall image information and operation information and transmits them to the detection module. The detection module performs disease category detection on the overall image information based on the detection strategy and the operation information to obtain the detection results; and transmits them to the business presentation module of the corresponding on-site device through the second communication module and the first communication module. The business presentation module displays the detection results.

[0068] The on-site device of this embodiment can be the on-site equipment of each user, and the server can be a professional detection server. It serves as a carrier of the detection technology of the technology manufacturer to centrally process the collected different train images and improve the image restoration accuracy and detection accuracy of the entire vehicle.

[0069] The image acquisition module of the on-site device can capture omnidirectional images of the train. The local industrial computer can then combine or mosaic the captured images to obtain comprehensive vehicle image information, i.e., appearance information, which can be sent to the server for defect detection. The on-site device of this embodiment can transmit the mosaicked vehicle image information, thereby improving transmission speed.

[0070] The image acquisition device of the field device in this embodiment can be installed in the following manner, for example: one image acquisition module is installed at the top of the track, one image acquisition module is installed at the side of the vehicle next to the track, and one image acquisition module is installed at the bottom of the vehicle on the track; one image acquisition module is installed in the track area to collect train speed-timetable, and one image acquisition module is installed in the track area to collect train number; Figure 3 As shown, each image acquisition module in this embodiment is a high-definition high-frequency linear array camera / high-definition high-speed 4K camera; alternatively, each image acquisition module takes pictures at a fixed frame rate (e.g., one thousand frames per second, etc.). The high-definition high-speed cameras on the trackside roof, side, and bottom of the vehicle involved here can use high-definition high-frequency linear array cameras, such as 2k or 4k cameras, depending on the vehicle model outline size and installation distance. During train inspection, the high-definition camera captures real-time images at a fixed frame rate. There are at least two linear array cameras on a single side of the vehicle on the trackside, and at least one linear array camera on the trackside roof and side of the vehicle.

[0071] It is particularly noted that the method of the present application is for photographing a dynamic high-speed running train, which may be an EMU or a high-speed train in current railway transportation.

[0072] In addition, in a specific implementation, the field device of this embodiment can be adjusted according to actual needs. For example, the field device can also include: a speed measurement unit, a fill light source, a trigger unit, etc. The field device of this embodiment can also be configured according to actual needs, which is not limited here and is only for example.

[0073] The speed measuring unit is used to measure the running speed of the train passing through the on-site device;

[0074] A fill light source is used to provide fill light when the image acquisition module acquires images; a fill light source can be set for each image acquisition module, or a fill light source that can be applied to multiple image acquisition modules can be set;

[0075] A trigger unit is used to start or shut down the image acquisition module of the field device. Each field device can be provided with a trigger unit, which can be installed in a position where it can sense the train. The time when it senses the train is earlier than the time when the image acquisition module collects the image.

[0076] When a maglev train passes through the trackside area, the trigger unit is activated, powering up the high-definition, high-speed cameras installed on the roof, sides, and underside of the train. These cameras begin capturing images and transmitting them to a local industrial computer for full-vehicle image restoration. The speed measurement unit also collects the train's speed and timetable, simultaneously triggering the vehicle identification system to identify the train's vehicle number. Once the train passes the high-definition, high-speed cameras, the trigger unit senses the situation and signals the shutdown of all high-definition, high-speed cameras and speed measurement units, triggering the local industrial computer to operate.

[0077] The local industrial computer can use the deep learning network for target recognition to identify feature points or feature objects in the collected image, and determine the position of the feature points or feature objects in the collected image based on the preset feature point or feature object pixel position and speed-timetable; then, according to the speed-timetable, uniform sampling and a small amount of redundant puzzle are used to restore the positions between the train feature points or feature objects to obtain the overall train image, and finally the vehicle information and the overall train image are transmitted to the server through the first communication module.

[0078] It is understood that the local industrial computer in this embodiment is specifically configured to receive a speed-timetable and determine train operation information; identify the head image of the train based on the image captured by the image acquisition module at the starting time point; and extract a frame of image from the image captured by the image acquisition module based on the speed v in the train operation information, the acquisition frequency P of the image acquisition module, and the width L of the acquisition frame; perform a concatenated splicing of the head image and the extracted image according to the image acquisition time point to generate overall image information of the train. At this point, the overall image information generated by the local industrial computer is redundant, meaning that the length of the overall image differs from the size of the train template in the detection module. This is primarily due to inaccurate speed measurement timing by the speed measurement system, resulting in relatively redundant overall image information generated by the local industrial computer. In other embodiments, the local industrial computer is specifically used to identify feature points (the head of the train, the tail of the train, and the middle key points) in the image captured by the image acquisition module, receive the speed-timetable, and determine the train operation information; according to the speed v in the train operation information, the acquisition frequency P of the image acquisition module, and the width L of the acquisition frame; based on the frame number step of (L / v) / 2, multiple frames of images are extracted from the feature point interval image, and redundant image splicing between feature points is performed; connected splicing is performed according to the image acquisition time point to generate the overall image information of the train; the multi-feature point recognition and re-jigsaw puzzle method of this embodiment reduces the accumulation of jigsaw puzzle errors caused by inaccurate speed measurement to a certain extent.

[0079] In other words, if speed measurement is accurate, keyframe images can be extracted and stitched based on the duration of L / v. However, due to inaccurate speed measurement, a duration of (L / v) / 2 is used to extract keyframe images for stitching. The applicable range of v is 5 to 30 km / h, and L depends on the accuracy of the camera / image acquisition module, which can be 1 mm or 0.5 mm.

[0080] The distributed detection system of this embodiment is designed based on the concept of "distributed collection and centralized detection". It uses a specialized detection server for centralized detection and processes the collected image data of different trains to improve the image restoration accuracy and detection accuracy of the entire vehicle.

[0081] like Figure 2 As shown, the local industrial computer of the field device in this embodiment can be specifically used to: identify feature points in each image, determine the position of the pixel to which the original train image belongs based on the pixel position and speed-time table to which each feature point belongs, and piece together all images into the overall image information of the train on the current track;

[0082] Identify partial information of the feature points of the image and determine the train operation information; the train operation information includes one or more of the following: train number, train speed, train model, detection unique serial number code, business instruction code, and user unit code.

[0083] In another optional implementation, the local industrial computer of the field device can identify feature points in each image based on a deep learning network, determine the pixel location of the original train image based on the pixel location and speed-time table of each feature point, and combine all images into an overall image of the train currently on the track;

[0084] Identify the speed-timetable and partial information of feature points in the image to determine the train operation information;

[0085] The train operation information includes one or more of the following: train number, train speed, train model, detection unique serial number code, business instruction code, and user unit code.

[0086] Typically, the train operation information uploaded to the server includes but is not limited to the user unit code, business instruction code and dynamic verification code, etc.; the train information includes but is not limited to the train model code, train car number code and the unique serial number code for the inspection of this trip.

[0087] In another optional implementation, the detection module of the server may be specifically configured to: select a train template to which the overall image information belongs based on the train operation information;

[0088] Based on the train operation information, selecting a train template to which the overall image information belongs;

[0089] The overall image information is used as a redundant image, and the redundant image is refined by means of a train template and a superpixel segmentation method or a semantic segmentation method, so that the redundant image is adjusted to a refined image corresponding to the overall image information;

[0090] Perform template matching between the refined image and the train template to obtain a detection result; or, match the refined image to a pre-trained deep network learning model to obtain a detection result output by the deep network learning model.

[0091] The detection results are classified into defects and suspected defects. Similarity between the suspected defects and the real defects and falsely detected defects in the expert experience database can be calculated to determine whether the suspected defects are real defects. It can be understood that compared to the template image, the redundant image is an image that is elongated horizontally (for the train's movement) and equal vertically.

[0092] The function of the detection module is to evenly spread the set 51200 or more (depending on the train length) cluster centers over the entire redundant image, and in the horizontal train area (such as Figure 6b dashed line range), on the cluster center line (such as Figure 6b The center solid line shows the sub-region size calibration, that is, Figure 6a The image template in the image is used for sub-region comparison and calibration. If the image features (texture, color, brightness, etc.) of the redundant image in a certain sub-region are similar to those of the image template, then the size calibration is performed directly; if the image information of the redundant image in a certain sub-region is very different, the sub-region in the vertical field (such as Figure 6b The image features are compared again until the sub-regions are similar and the size is calibrated.

[0093] The superpixel segmentation and correction algorithm has high requirements for server configuration and is also suitable for centralized detection strategies to save computing resources.

[0094] To better illustrate, Figure 2 As shown, in actual application, the detection module in the server may specifically include:

[0095] A vehicle type screening unit, based on the train operation information, determines the train type An to which the overall image information currently sent by the on-site device belongs and the train template to which the train type An belongs;

[0096] The An model appearance detection unit performs matching based on the train template to which the train model An belongs and the overall image information to obtain a detection result.

[0097] The detection module in the server of this embodiment can use the secondary correction algorithm of the whole vehicle image to correct the whole vehicle image, thereby improving the accuracy of the whole vehicle image restoration; at the same time, it has the ability of algorithm iteration and experience summary, thereby improving the long-term detection accuracy of the train appearance image detection system.

[0098] It should be noted that Figure 2 The server used for professional detection shown in may also include: an expert experience module;

[0099] The expert experience module of this embodiment is used to regularly collect true defect detection results and false positive results for the same vehicle model as an expert experience database. This expert experience database is used to periodically correct the experience values ​​used by the detection module. Furthermore, the expert experience database is used to provide reference information for true defect detection results and false positive results used by the detection module in similarity comparison testing. In this embodiment, one expert experience database can be assigned to each vehicle model.

[0100] In other words, the detection module processes the image information of the corresponding vehicle model and outputs the disease category and image coordinates of the train appearance; the expert experience module mainly accumulates the disease occurrence category and its corresponding life cycle of each vehicle model, provides users with experience guidance for major inspections, and provides data support for decisions on subsequent detection algorithm upgrades.

[0101] The detection module can correct the redundant train images uploaded by the on-site device. The method is to perform superpixel segmentation or semantic segmentation on the redundant train image and the template image of the vehicle, and then fit the sub-region size of the image data of the two in the direction of the train's movement to obtain a corrected image; the corrected image is then used through template matching and deep learning to identify the overall appearance defects of the train (such as cracks, damage, missing parts, etc.).

[0102] During the inspection process, the expert experience module accumulates actual inspection results for a specific vehicle type. Once sufficient experience is accumulated, it adjusts the empirical values ​​used to determine certain inspection results, such as the confidence threshold for target recognition and the difference threshold for template matching. Maintenance experience accumulates the lifecycle and frequency of failures in various train components, providing preventive maintenance experience within the user's maintenance plan for trains of the same series or model.

[0103] The above-mentioned system detection and analysis process is implemented in the professional servers of technology manufacturers, reducing the hardware cost of the target detection server for each detection project; facilitating the update and iteration of detection technology by technology manufacturers, and improving the detection accuracy of image detection projects for various vehicle models; providing full life cycle analysis and prediction, facilitating the establishment of big data for rail train appearance image detection; and centralizing and efficiently utilizing high-performance professional detection servers to avoid waste of computing resources.

[0104] Professional detection servers can correct train images again and detect targets, reducing the software burden on on-site industrial computers and greatly reducing dependence on on-site high-precision speed measurement systems.

[0105] The business cooperation between users and technology manufacturers' projects is closer, the long-term user experience is enhanced, and the long-term technology accumulation speed of technology manufacturers is accelerated.

[0106] In addition, if Figure 4 As shown, the service presentation module provided in this embodiment may include:

[0107] The alarm unit is used to send out an alarm signal when the detection result is a serious damage level. The alarm unit can be a sound alarm, a light signal alarm, etc., to effectively remind relevant personnel to confirm and repair offline in a timely manner.

[0108] An analysis unit is used to analyze and process the test results within a preset time period to obtain report information and life cycle prediction information;

[0109] The display unit is used to display each test result.

[0110] It is understandable that the analysis unit and the presentation unit may be components in the Web business system. Figure 4 The figure shows the alarm unit and the web business system. The web business system includes an API for receiving test results; general functional modules (such as user management, report generation, history query, and statistical analysis); and a page for users to review and confirm test results. This embodiment does not limit the interface display functions of the web business system and can be configured according to actual needs.

[0111] like Figure 5 As shown, the second communication module and the first communication module in the field device and the server may include: a firewall unit and a transmission unit;

[0112] Among them, the firewall unit is used to encrypt and decrypt data sent to the external network;

[0113] The transmission unit is used for transmission via optical communication or 5G communication. The communication module can include high-speed communication technologies such as optical fiber communication and 5G technology to improve the speed of long-distance communication and ensure the rapid and real-time transmission of the overall image of the train appearance.

[0114] The above-mentioned on-site device and the professional detection server cooperate to realize the rapid detection and analysis of the appearance image of the dynamic train at the lowest cost and with high detection accuracy.

[0115] It should be noted that, in the claims, any reference signs placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention may be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In claims enumerating several means, several of these means may be embodied by one and the same hardware. The use of the words first, second, third etc. is for convenience only and does not indicate any order. These words may be understood as part of the component name.

[0116] In addition, it should be noted that, in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0117] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments after learning the basic creative concept. Therefore, the claims should be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0118] Obviously, those skilled in the art may make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if such modifications and variations fall within the scope of the claims and their equivalents, the present invention shall also include such modifications and variations.

Claims

1. A rail transit train image collection and distribution detection system, characterized in that: include: multiple field devices and a server; Each on-site device includes: multiple image acquisition modules, a local industrial computer, a first communication module and a business presentation module; the server includes: a second communication module and a detection module; Multiple image acquisition modules are respectively installed in designated areas beside the designated track to transmit the acquired images to the local industrial computer; all image acquisition modules are electrically connected to the local industrial computer; The local industrial computer performs a first puzzle based on all images and vehicle speed information transmitted by the connected image acquisition module to generate overall image information of the train on the current track and operation information of the train, and sends the image information and operation information to the server through the first communication module; The second communication module receives the overall image information and operation information and transmits them to the detection module. The detection module performs disease category detection on the overall image information based on the detection strategy and the operation information to obtain the detection results; and transmits them to the business presentation module of the corresponding on-site device through the second communication module and the first communication module. The business presentation module displays the detection results.

2. The detection system according to claim 1, characterized in that One image acquisition module is installed on the top of the track, one image acquisition module is installed on the side of the vehicle next to the track, and one image acquisition module is installed on the bottom of the vehicle on the track; one image acquisition module is installed in the track area to collect train speed and timetable, and one image acquisition module is installed in the track area to collect train number; Each image acquisition module is a high-definition high-frequency line array camera / high-definition high-speed 4K camera; Each image acquisition module takes pictures at a fixed frame rate.

3. The detection system according to claim 1, characterized in that The local industrial computer is specifically used for Identify the feature points in each image, determine the location of the pixels in the original train image based on the pixel location and speed-time table of each feature point, and piece together all the images to form the overall image information of the train on the current track; Identify partial information of the feature points of the image and determine the train operation information; or, The local industrial computer uses a deep learning network to identify feature points in each image. Based on the pixel location and speed-time table of each feature point, it determines the location of the pixel in the original train image and then pieces together all the images to create a comprehensive image of the train currently on the track. Identify the speed-timetable and partial information of feature points in the image to determine the train operation information; The train operation information includes one or more of the following: train number, train speed, train model, detection unique serial number code, business instruction code, and user unit code.

4. The detection system according to claim 1, characterized in that The local industrial computer is specifically used for The image acquisition module identifies the feature points in the image, receives the speed-timetable, and determines the train operation information; According to the speed v in the train operation information and the width L of the acquisition frame, multiple frames are extracted from the feature point interval image based on the frame number step of (L / v) / 2, and redundant image splicing between feature points is performed; the images are connected and spliced ​​according to the image acquisition time point to generate the overall image information of the train.

5. The detection system according to claim 1 or 4, characterized in that: The server also includes: an expert experience module; The expert experience module is used to regularly collect the real disease detection results and false disease detection results of the same vehicle model as the expert experience database; The expert experience database is used to periodically correct the experience values ​​used by the detection module; Furthermore, the expert experience database is used to provide reference information of real disease detection results and false disease detection results used by the detection module in similarity comparison detection; The expert experience database has one expert experience database for each vehicle model.

6. The detection system according to claim 5, characterized in that: The detection module is specifically used for Based on the train operation information, selecting a train template to which the overall image information belongs; The overall image information is used as a redundant image, and the redundant image is refined by means of a train template and a superpixel segmentation method or a semantic segmentation method, so that the redundant image is adjusted to a refined image corresponding to the overall image information; Perform template matching on the refined image and the train template to obtain a detection result; or, match the refined image to a pre-trained deep network learning model to obtain a detection result output by the deep network learning model; The detection results are classified into diseases and suspected diseases, and the similarity between the suspected diseases and the real diseases and misdetected diseases in the expert experience database is calculated to determine whether the suspected diseases are real diseases.

7. The detection system according to claim 1, characterized in that The business presentation module includes: An alarm unit, configured to send out an alarm signal when the detection result is a detection result of a serious damage level; An analysis unit is used to analyze and process the test results within a preset time period to obtain report information and life cycle prediction information; The display unit is used to display each test result.

8. The detection system according to claim 1, characterized in that The detection module includes: A vehicle type screening unit, based on the train operation information, determines the train type An to which the overall image information currently sent by the on-site device belongs and the train template to which the train type An belongs; The An model appearance detection unit performs matching based on the train template to which the train model An belongs and the overall image information to obtain a detection result.

9. The detection system according to claim 1, characterized in that: The field device further comprises: Speed ​​measuring unit, used to measure the running speed of the train passing through the on-site device; A fill light source, used for providing fill light when the image acquisition module is acquiring images; A trigger unit is used to start the field device.

10. The detection system according to claim 1, characterized in that: The second communication module and the first communication module both include: Firewall unit, used to encrypt and decrypt data sent to the external network; A transmission unit is used for transmission by means of optical communication or by means of 5G communication.

Images