Processing system for detecting bottom of dynamic and static rail train

Abstract

The embodiment of the invention relates to a processing system for detecting the bottom of a dynamic and static rail train. The system comprises a static acquisition equipment group, a dynamic acquisition equipment group and a remote platform, the static group collects visible light and ultraviolet light images through inspection of a vehicle bottom robot, performs oil leakage fault detection according to an ultraviolet light image set, and performs bolt loosening and part damage fault detection according to a visible light image set based on a visual fault recognition model; the dynamic group generates a vehicle bottom full image through a line-scan digital camera, and oil leakage fault detection is carried out according to the vehicle bottom full image; and the remote platform stores the static and dynamic detection reports received each time into a corresponding train detection library, and regularly upgrades and updates the visual fault identification model. The method can improve the detection precision, reduce the detection cost, improve the detection efficiency, improve the maintenance frequency of each vehicle, and make up the defects of a conventional scheme.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a processing system for undercarriage inspection of dynamic and static rail trains. Background Technology

[0002] Lubricating oil leaks in critical components of railcars, such as gearboxes and hydraulic systems, are common malfunctions threatening operational safety, necessitating regular or irregular inspections of the train's undercarriage. The traditional method involves parking the train in a designated depot and conducting a comprehensive manual inspection of the undercarriage, then determining whether repairs are needed based on the inspection results. This traditional method is limited by labor efficiency and depot capacity, making it difficult to shorten the maintenance cycle for individual trains and, consequently, increase the frequency of maintenance for all trains.

[0003] With the development and application of artificial intelligence technology, the use of visual target recognition models for detecting oil leaks under vehicles has been gradually applied to under-vehicle inspection scenarios. While this AI-powered under-vehicle inspection solution can indeed shorten the inspection cycle for a single vehicle, it also has some shortcomings: 1) Water and lubricating oil have highly similar visual characteristics, and traditional visual target recognition models have limited accuracy in distinguishing between water and oil areas, leading to decreased detection accuracy; 2) Under-vehicle inspection solutions using AI models are mostly single-modal (only detecting lubricating oil leaks). Detection of other faults (such as loose bolts or broken parts) still requires manual intervention using multiple sets of auxiliary equipment, which increases inspection costs and reduces efficiency; 3) Under-vehicle inspection solutions using AI models are mostly used in static inspection scenarios, lacking real-time detection methods for dynamically moving vehicles. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a processing system for undercarriage inspection of static and dynamic railcars. This system includes: a static acquisition equipment group, a dynamic acquisition equipment group, and a remote platform. The static acquisition equipment group uses an undercarriage robot to collect visible light and ultraviolet (UV) filtered images of all designated component locations on the undercarriage of a static train parked on the tracks inside the depot. It then uses a dynamic brightness threshold method to detect oil leaks based on the UV image set and a visual fault recognition model to detect bolt loosening and component damage based on the visible light image set. The dynamic acquisition equipment group uses a line scan camera to collect UV filtered images of the undercarriage of a moving train and uses a dynamic brightness threshold method to detect oil leaks based on the full undercarriage image. The remote platform stores the dynamic and static inspection reports for each train and periodically upgrades and trains the visual fault recognition model used by the undercarriage robot. The ultraviolet light image detection method of the present invention can improve the detection accuracy of lubricating oil leakage areas. The static multimodal detection method of the present invention can further compress the single vehicle maintenance cycle, reduce detection costs, improve detection efficiency, and increase the maintenance frequency of each vehicle. The dynamic detection mechanism of the present invention can make up for the defects of conventional solutions and increase the detection frequency of each vehicle.

[0005] To achieve the above objectives, embodiments of the present invention provide a processing system for undercarriage inspection of static and dynamic rail trains, the system comprising: a static acquisition equipment group, a dynamic acquisition equipment group, and a remote platform; The remote platform is connected to the static acquisition equipment group and the dynamic acquisition equipment group respectively; the lubricating oil used in all undercarriage components of any train detected by the static acquisition equipment group and the dynamic acquisition equipment group contains fluorescent agent, and the lubricating oil is delivered to each component through a closed hydraulic pipeline; The static data acquisition equipment group is used to collect visible light and ultraviolet light filtered images of all designated component locations on the undercarriage of a static train parked on the tracks inside the garage through an undercarriage robot inspection, obtaining corresponding visible light image sets and ultraviolet light image sets; and to perform oil leakage fault detection based on the ultraviolet image sets using a dynamic brightness threshold method to obtain a corresponding first detection report; and to perform bolt loosening and component damage fault detection based on the visible light image sets using a visual fault recognition model to obtain a corresponding second detection report; and to generate a corresponding static detection report based on the current train's basic information and the first and second detection reports, and send it to the remote platform. The linear array camera of the dynamic acquisition device group is fixedly installed at the bottom of the designated track. The dynamic acquisition device group is used to acquire data from the ultraviolet filtered full-body image of the undercarriage of a moving train passing through the current track position to obtain the corresponding first full-body image of the undercarriage. Based on the dynamic brightness threshold method, the group performs oil leakage fault detection based on the first full-body image of the undercarriage and obtains the corresponding dynamic detection report, which is then sent to the remote platform. The remote platform is used to store multiple train inspection databases; each train inspection database corresponds one-to-one with a train; the train inspection database is used to store all the static inspection reports and dynamic inspection reports of the corresponding train; The remote platform is also used to store each received static or dynamic test report into the corresponding train test database; The remote platform is also used to periodically upgrade and train the visual fault recognition model and synchronize the latest trained model to all the static acquisition device groups.

[0006] Preferably, the static acquisition equipment group includes an under-vehicle inspection robot, a six-degree-of-freedom robotic arm, a light source switching component, a filter switching component, and a first camera; the light source switching component has two built-in ambient light sources for the first camera, namely a first visible light source and a first ultraviolet light source; the filter switching component has two built-in lens filters for the first camera, namely a first visible filter and a first ultraviolet filter; the first visible filter is used to transmit light in a preset visible light band and filter light in any band outside the visible light band; the first ultraviolet filter is used to transmit light in a preset first ultraviolet light band and filter light in any band outside the first ultraviolet light band; the first camera is an area array camera; The under-vehicle inspection robot is connected to the remote platform via wireless communication and communicates with the six-degree-of-freedom robotic arm, the light source switching component, the filter switching component, and the first camera via wired or wireless data communication. The six-degree-of-freedom robotic arm is mounted on the under-vehicle inspection robot and can rotate at any angle under the scheduling and control of the under-vehicle inspection robot. The light source switching component, the filter switching component, and the first camera are installed at three fixed positions at the end of the six-degree-of-freedom robotic arm. The undercarriage inspection robot has a built-in visual fault recognition model; the visual fault recognition model is based on a type of visual target recognition model; the visual target recognition model includes the YOLO series models; the visual fault recognition model is used to perform target detection and classification recognition processing on the input image and output a corresponding set of target detection boxes; when the set of target detection boxes is not empty, it consists of one or more target detection boxes; the target detection box includes the coordinates of the center point of the detection box, the height of the detection box, the width of the detection box, and the target type; the target type includes loose bolts and broken parts. The dynamic acquisition equipment group includes a trackside intelligent control device, a second ultraviolet light source, a second ultraviolet filter, and a second camera; the second ultraviolet filter is used to transmit light in a preset second ultraviolet light band and filter light in any band other than the second ultraviolet light band; the second camera is a line scan camera. The trackside intelligent control device is connected to the remote platform via wireless communication and communicates with the second ultraviolet light source and the second camera via wired or wireless data communication. The trackside intelligent control device is fixedly installed at a designated trackside location on the designated train track. The second ultraviolet light source and the second camera are fixedly installed at two designated bottom locations on the designated train track. The linear scan direction of the second camera is perpendicular to the train's direction of travel on the designated train track. The second ultraviolet filter is fixedly installed at the front end of the lens of the second camera.

[0007] Preferably, the static acquisition equipment group is specifically used to acquire data from the visible light and ultraviolet light filtered images of all designated component locations on the undercarriage of a static train parked on the tracks inside the garage through the inspection by the undercarriage robot, to obtain the corresponding visible light image set and ultraviolet light image set: The undercarriage inspection robot pauses when dispatched by the operator to the initial inspection point of any static train and receives the first train identifier and first train model input by the operator. Based on the first train model, it queries the built-in undercarriage inspection point map library to obtain the corresponding first inspection point map. Based on the first inspection point map, it plans the inspection path to obtain the corresponding first inspection path and initiates the current inspection process based on the first inspection path. During the current inspection, the built-in LiDAR or visual SLAM navigation system performs motion navigation according to the first inspection path. During the current inspection, it pauses at the trajectory point corresponding to each first inspection point on the first inspection point map. Each time the trajectory is paused, the six-degree-of-freedom robotic arm is first dispatched to send the first camera to the shooting position corresponding to the current undercarriage point, and then the light source switching component is dispatched to turn on the first visible light source, and the filter switching group is dispatched. The process involves the first camera switching to the first visible light filter, then scheduling the first camera to capture a filtered image to obtain a corresponding first visible light image. Next, the light source switching component is scheduled to turn off the first visible light source and turn on the first ultraviolet light source; the filter switching component is scheduled to switch the first ultraviolet filter; then the first camera is scheduled to capture a filtered image to obtain a corresponding first ultraviolet light image; finally, the light source switching component is scheduled to turn off the first ultraviolet light source. First point data is formed from the first point marker corresponding to the current inspection point and the first visible light image; second point data is formed from the current first point marker and the first ultraviolet light image. The current inspection process ends when the end trajectory point of the current inspection path is reached. At the end of the current inspection process, all the obtained first point data forms the visible light image set, and all the obtained second point data forms the ultraviolet light image set. The vehicle undercarriage inspection point map library includes multiple first point map records; each first point map record includes a first vehicle model and a first inspection point map; each first inspection point map includes multiple first inspection points; each first inspection point corresponds to a designated component location on the undercarriage of the current vehicle model; each first inspection point includes a first point identifier and a first point relative position; the first point relative position includes a first lateral displacement and a first longitudinal displacement; the first lateral displacement and the first longitudinal displacement are the lateral and longitudinal relative displacement distances between the previous inspection point and the initial inspection point, respectively.

[0008] Preferably, the static acquisition device group is specifically used when the oil leak fault detection based on the ultraviolet light image set using the dynamic brightness threshold method is performed to obtain the corresponding first detection report: The undercarriage inspection robot uses each of the first ultraviolet images in the ultraviolet image set as the current ultraviolet image; The brightness of each pixel in the current ultraviolet light image is identified to obtain the corresponding pixel brightness; the average value of all pixel brightness is used as the corresponding reference brightness; the product of a preset first multiple and the reference brightness is used as the corresponding current brightness threshold, where the first multiple is a positive integer greater than 2; all pixels with a brightness not less than the current brightness threshold are recorded as fluorescent points; the total number of fluorescent points is counted to obtain a total number N1; and it is determined whether the total number N1 is zero. If N1=0, then set the corresponding first point detection box set to empty; If N1 > 0, then all the fluorescence points in the current ultraviolet image are clustered to obtain one or more corresponding first point sets; and it is identified whether the total number of fluorescence points in each first point set is less than a preset first total threshold. If so, the current point set is deleted; and it is identified whether the total number of points in the remaining first point sets is zero. If so, the corresponding first point detection box set is set to empty; if not, a corresponding first detection box is formed by the image coordinates of the four boundary points of the remaining first point sets, and the corresponding first point detection box set is formed by all the obtained first detection boxes; wherein, at least one of the point distances between any point in the first point set and all other points in the current point set is less than a preset first point distance threshold. The current ultraviolet light image and its corresponding first point identifier and the set of first point detection boxes constitute the corresponding first point detection record; The first detection report is composed of all the detection records of the first point obtained.

[0009] Preferably, the static acquisition device group is specifically used when the second detection report is obtained by detecting bolt loosening and component damage based on the visible light image set according to the visual fault recognition model: The undercarriage inspection robot uses each of the first visible light images in the visible light image set as the current visible light image; The current visible light image is used as the current model input image and fed into the visual fault recognition model for target detection and classification to obtain the corresponding target detection box set; and the current target detection box set is used as the corresponding second point detection box set. The current visible light image and its corresponding first point identifier and second point detection box set constitute the corresponding second point detection record; The second detection report is composed of all the detection records of the second points.

[0010] Preferably, the static data acquisition device group is specifically used when the corresponding static inspection report is generated based on the current train's basic information and the first and second inspection reports and sent to the remote platform: The undercarriage inspection robot consists of the start and end times of the current inspection process, forming a corresponding first inspection period; and a corresponding static inspection report, consisting of the first inspection period, the first train identifier, the first train model, the first inspection point map, the first inspection report, and the second inspection report, is sent to the remote platform.

[0011] Preferably, the dynamic acquisition device group is specifically used to acquire data from the ultraviolet-filtered full-body image of the undercarriage of a train passing through the current track position to obtain the corresponding first full-body image of the undercarriage: The trackside intelligent control device identifies the time period of each train passing through the current track position based on the built-in train passage sensing module; at the beginning of the current train passage time, it turns on the second ultraviolet light source for continuous illumination and starts the second camera to capture filtered images; at the end of the current train passage time, it turns off the second ultraviolet light source and stops the second camera's shooting operation, and obtains the corresponding first line array image sequence from the second camera; and according to the line array image stitching principle, it stitches the undercarriage length image based on the first line array image sequence and uses the stitched undercarriage length image as the corresponding first undercarriage full image; The first linear array sequence is composed of multiple first linear arrays arranged in chronological order, with all first linear arrays having the same image height and width. The linear array stitching principle is to stitch all the linear arrays in the sequence in chronological order with time as the vertical stitching direction to obtain a vertical long image. The image width of the first full-view image of the vehicle's underside is the same as the image width of the first linear array, and the image height of the first full-view image of the vehicle's underside is the sum of the image heights of all the first linear arrays.

[0012] Preferably, the dynamic acquisition device group is specifically used to send the corresponding dynamic detection report obtained from the oil leak fault detection based on the first full-view image of the vehicle underside using the dynamic brightness threshold method to the remote platform: The trackside intelligent control device identifies the brightness of each pixel in the full image of the first vehicle underside to obtain the corresponding pixel brightness; and uses the average value of all pixel brightness as the corresponding reference brightness; and uses the product of a preset second multiple and the reference brightness as the corresponding current brightness threshold, where the second multiple is a positive integer greater than 2; and records all pixels whose pixel brightness is not less than the current brightness threshold as fluorescent points; and counts the total number of fluorescent points to obtain a total number N2; and identifies whether the total number N2 is zero. If N2=0, then set the corresponding undercarriage detection box set to empty; If N2 > 0, then all the fluorescent points in the first vehicle underside full image are clustered to obtain one or more corresponding second point sets; and it is identified whether the total number of fluorescent points in each second point set is less than a preset second total threshold. If so, the current point set is deleted; and it is identified whether the total number of points in the remaining second point sets is zero. If so, the corresponding vehicle underside detection box set is set to empty; if not, a corresponding second detection box is formed by the image coordinates of the four boundary points of each remaining second point set, and the corresponding vehicle underside detection box set is formed by all the obtained second detection boxes; wherein, at least one of the point distances between any point in the second point set and all other points in the current point set is less than a preset second point distance threshold. The current equipment group's corresponding track marker and track installation position are used as the corresponding first track marker and first track position; the current train passage time corresponding to the first undercarriage full map is used as the corresponding first detection time period; and the dynamic detection report composed of the first track marker, the first track position, the first detection time period, and the undercarriage detection frame set is sent to the remote platform.

[0013] Preferably, the remote platform is specifically used when storing each received static test report or dynamic test report into the corresponding train test database; Upon receiving the static inspection report, the corresponding first train identifier is extracted from the current static inspection report; and the current static inspection report is stored in the train inspection database corresponding to the current first train identifier. Upon receiving the dynamic detection report, the corresponding first track identifier, first track position, and first detection period are extracted from the current dynamic detection report; and through a preset nationwide train scheduling query interface, the unique identifier of the train that travels along the track corresponding to the first track identifier and passes through the first track position during the first detection period is queried, and the query result is used as the corresponding current train identifier; and the current dynamic detection report is stored in the train detection database corresponding to the current train identifier.

[0014] Preferably, the remote platform is specifically used when periodically upgrading and training the visual fault recognition model and synchronizing the latest trained model to all the static acquisition device groups: On the platform side, large-scale big data collection is performed on under-vehicle inspection images of all operating models on the entire network under various lighting conditions to obtain the current raw image set; and a corresponding model training dataset is constructed based on the raw image set; and the visual fault recognition model is trained once based on the model training dataset; and at the end of this round of training, the latest visual fault recognition model is synchronized to the under-vehicle inspection robots of all the static acquisition equipment groups. The original image set includes multiple undercarriage inspection images; each undercarriage inspection image is a visible light image; in the original image set, some undercarriage inspection images correspond to areas with no loose bolts or damaged components, while others correspond to areas with one or more loose bolts and / or damaged components; the undercarriage inspection images corresponding to areas without loose bolts or damaged components do not have detection box labels; the undercarriage inspection images corresponding to areas with loose bolts or damaged components have one or more corresponding detection box labels, and the detection box labels correspond one-to-one with the locations of loose bolts or damaged components in the current undercarriage area.

[0015] This invention provides a processing system for inspecting the undercarriage of static and dynamic railcars. As described above, the system includes: a static acquisition equipment group, a dynamic acquisition equipment group, and a remote platform. The static acquisition equipment group uses an undercarriage robot to collect visible light and ultraviolet (UV) filtered images of all designated component locations on the undercarriage of a static train parked on the tracks inside the depot. It then uses a dynamic brightness threshold method to detect oil leaks based on the UV image set and a visual fault recognition model to detect bolt loosening and component damage based on the visible light image set. The dynamic acquisition equipment group uses a line scan camera to collect UV filtered images of the undercarriage of a moving train and uses a dynamic brightness threshold method to detect oil leaks based on the full undercarriage image. The remote platform stores the dynamic and static inspection reports for each train and periodically upgrades and trains the visual fault recognition model used by the undercarriage robot. The ultraviolet light image detection method of this invention improves the detection accuracy of lubricating oil leakage areas, the static multimodal detection method compresses the maintenance cycle of a single vehicle, reduces the inspection cost of the whole vehicle, improves the detection efficiency, and increases the maintenance frequency of each vehicle, and the dynamic detection mechanism makes up for the defects of conventional solutions and increases the inspection frequency of each vehicle. Attached Figure Description

[0016] Figure 1 This is a module structure diagram of a processing system for undercarriage inspection of dynamic and static rail trains, provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0018] This invention provides a processing system for undercarriage inspection of dynamic and static rail trains, such as... Figure 1 The module structure diagram of a processing system for undercarriage inspection of static and dynamic rail trains provided in an embodiment of the present invention is shown. It mainly includes: a static acquisition device group 1, a dynamic acquisition device group 2, and a remote platform 3. The remote platform 3 is connected to both the static acquisition device group 1 and the dynamic acquisition device group 2.

[0019] It should be noted that, in this embodiment of the invention, fluorescent agents, such as coumarin derivatives and rhodamine B, are added to the lubricating oil used by all undercarriage components of any train under operating conditions, as detected by the static acquisition device group 1 and the dynamic acquisition device group 2. The lubricating oil used by all undercarriage components of the railcar is delivered to each component through closed hydraulic pipelines. A lubricating oil leak occurs when an oil leak point appears in the pipeline and / or on the component. After adding fluorescent agents to the lubricating oil, if an oil leak area exists within the imaging range during ultraviolet light imaging, a corresponding bright area will appear on the ultraviolet image. Conversely, water accumulation / stains adhering to the pipeline and / or components will not produce a fluorescent effect and will not appear as bright areas on the ultraviolet image. This allows for a clear distinction between water and oil areas.

[0020] (a) Static data acquisition equipment group 1: The static acquisition device group 1 of this invention includes a vehicle under-vehicle inspection robot 11, a six-degree-of-freedom robotic arm 12, a light source switching component 13, a filter switching component 14, and a first camera 15.

[0021] The undercarriage inspection robot 11 is connected to the remote platform 3 via wireless communication, and communicates with the six-degree-of-freedom robotic arm 12, the light source switching component 13, the filter switching component 14 and the first camera 15 via wired or wireless data communication.

[0022] The undercarriage inspection robot 11 has a built-in visual fault recognition model. This embodiment of the invention uses a visual target recognition model based on a class of visual target recognition models; this model includes at least the YOLO series models. The visual fault recognition model performs target detection and classification on the input image and outputs a corresponding set of target detection boxes; wherein, when the target detection box set is not empty, it consists of one or more target detection boxes; each target detection box includes the coordinates of its center point, its height, its width, and the target type; target types include loose bolts and damaged components.

[0023] A six-degree-of-freedom robotic arm 12 is mounted on an undercarriage inspection robot 11 and is controlled by the undercarriage inspection robot 11 to rotate at any angle. A light source switching assembly 13, a filter switching assembly 14, and a first camera 15 are installed at three fixed positions at the end of the six-degree-of-freedom robotic arm 12.

[0024] The light source switching component 13 has two built-in ambient light sources for the first camera 15, namely the first visible light source 131 and the first ultraviolet light source 132.

[0025] The filter switching assembly 14 has two built-in lens filters for the first camera 15: a first visible light filter 141 and a first ultraviolet filter 142. The first visible light filter 141 is used to transmit light within a preset visible light band and filter light of any band outside the visible light band; the first ultraviolet filter 142 is used to transmit light within a preset first ultraviolet light band and filter light of any band outside the first ultraviolet light band.

[0026] Here, the visible light band and the first ultraviolet light band in this embodiment of the invention are two pre-set light band ranges, which can be customized based on actual application needs.

[0027] The first camera 15 is an area array camera.

[0028] Static acquisition equipment group 1 is used to collect visible light and ultraviolet light filtered images of all designated component locations on the static train undercarriage parked on the tracks inside the garage through undercarriage robot inspection, and obtain corresponding visible light image sets and ultraviolet light image sets; and to detect oil leakage faults based on the ultraviolet light image sets using the dynamic brightness threshold method, and obtain the corresponding first inspection report; and to detect bolt loosening and component damage faults based on the visible light image sets using the visual fault recognition model, and obtain the corresponding second inspection report; and to generate the corresponding static inspection report based on the basic information of the current train and the first and second inspection reports, and send it to the remote platform 3.

[0029] In one specific implementation of this invention, the static acquisition device group 1 is specifically used to acquire data from the visible light and ultraviolet light filtered images of all designated component locations on the undercarriage of a static train parked on the tracks inside the garage through an undercarriage robot inspection, to obtain the corresponding visible light image set and ultraviolet light image set: Step A1: When the undercarriage inspection robot 11 is dispatched by the operator to the initial inspection point of any static train, it pauses and receives the first train identifier and the first train model input by the operator. Step A2, and obtain the corresponding first inspection point map by querying the built-in undercarriage inspection point map library based on the first train model; Here, the vehicle undercarriage inspection point map library of this invention includes multiple first point map records; each first point map record includes a first vehicle model and a first inspection point map; each first inspection point map includes multiple first inspection points; each first inspection point corresponds to a designated component location point on the undercarriage of the current vehicle model; each first inspection point includes a first point identifier and a first point relative position; the first point relative position includes a first lateral displacement and a first longitudinal displacement; the first lateral displacement and the first longitudinal displacement are respectively the lateral and longitudinal relative displacement distances between the previous inspection point and the initial inspection point; The query based on the first train model is actually the extraction of the first inspection point map of the first inspection point map record that matches the first train model in the first inspection point map map library, and it is used as the query result for the current time. Step A3, and based on the first inspection point location map, perform inspection path planning to obtain the corresponding first inspection path; Step A4, and start the inspection process based on the first inspection path; Step A5, and during the current inspection, the built-in LiDAR or visual SLAM navigation system will perform motion navigation according to the first inspection path; Step A6, and during the current inspection process, pause at the trajectory point position corresponding to each first inspection point on the first inspection point location map; Step A7: During each trajectory pause, first, the six-degree-of-freedom robotic arm 12 is dispatched to send the first camera 15 to the shooting position corresponding to the current vehicle under-point. Then, the light source switching component 13 is dispatched to turn on the first visible light source 131, and the filter switching component 14 is dispatched to switch the first visible filter 141 for the first camera 15. Then, the first camera 15 is dispatched to capture the filtered image to obtain the corresponding first visible light image. Then, the light source switching component 13 is dispatched to turn off the first visible light source 131 and turn on the first ultraviolet light source 132, and the filter switching component 14 is dispatched to switch the first ultraviolet filter 142. Then, the first camera 15 is dispatched to capture the filtered image to obtain the corresponding first ultraviolet light image. Then, the light source switching component 13 is dispatched to turn off the first ultraviolet light source 132. The first point data is composed of the first point marker and the first visible light image corresponding to the current inspection point, and the second point data is composed of the current first point marker and the first ultraviolet light image. Step A8, and end the current inspection process when the end trajectory point of the current inspection path is reached; and at the end of the current inspection process, a visible light image set is composed of all the first point data obtained, and an ultraviolet light image set is composed of all the second point data obtained.

[0030] In another specific implementation of this invention, the static acquisition device group 1 is specifically used to obtain the corresponding first detection report when performing oil leak fault detection based on the ultraviolet light image set using the dynamic brightness threshold method: Step B1: The undercarriage inspection robot 11 uses each of the first ultraviolet images in the ultraviolet image set as the current ultraviolet image. Step B2 involves identifying the brightness of each pixel in the current ultraviolet image to obtain the corresponding pixel brightness; using the average value of all pixel brightness as the corresponding reference brightness; using the product of a preset first multiple and the reference brightness as the corresponding current brightness threshold, where the first multiple is a positive integer greater than 2; recording all pixels with a brightness not less than the current brightness threshold as fluorescent points; counting the total number of fluorescent points to obtain the total number N1; and identifying whether the total number N1 is zero. Step B3: If N1=0, then set the corresponding first point detection box set to empty; Step B4: If N1 > 0, cluster all fluorescence points in the current ultraviolet image to obtain one or more corresponding first point sets; identify whether the total number of fluorescence points in each first point set is less than a preset first total threshold; if so, delete the current point set; identify whether the total number of points in the remaining first point sets is zero; if so, set the corresponding first point detection box set to empty; if not, form a corresponding first detection box by the image coordinates of the four boundary points of the remaining first point sets, and form the corresponding first point detection box set by all the obtained first detection boxes. Here, in this embodiment of the invention, at least one of the point distances between any point in the first point set and all other points in the current point set is less than a preset first point distance threshold. In this embodiment of the invention, the first total threshold and the first point spacing threshold are two preset threshold parameters that can be customized based on actual application requirements. Step B5, and the corresponding first point detection record is composed of the current ultraviolet light image and its corresponding first point identifier and first point detection box set; Step B6, and the corresponding first detection report is composed of all the first point detection records obtained.

[0031] In another specific implementation of this invention, the static acquisition device group 1 is specifically used to obtain a corresponding second detection report when detecting bolt loosening and component damage based on a visible light image set using a visual fault recognition model: Step C1: The undercarriage inspection robot 11 uses each of the first visible light images in the visible light image set as the current visible light image. Step C2: The current visible light image is used as the current model input image and fed into the visual fault recognition model for target detection and classification to obtain the corresponding target detection box set; and the current target detection box set is used as the corresponding second point detection box set. Step C3, and the corresponding second point detection record is composed of the current visible light image and its corresponding first point identifier and second point detection box set; Step C4, and the corresponding second detection report is composed of all the second point detection records obtained.

[0032] In another specific implementation of this invention, the static data acquisition device group 1 is specifically used to generate a corresponding static detection report based on the basic information of the current train and the first and second detection reports and send it to the remote platform 3: The undercarriage inspection robot 11 consists of the start and end times of the current inspection process, forming the corresponding first inspection period; and it sends the corresponding static inspection report, consisting of the first inspection period, the first train identification, the first train model, the first inspection point map, the first inspection report, and the second inspection report, to the remote platform 3.

[0033] (ii) Dynamic data acquisition equipment group 2: The dynamic acquisition device group 2 of this embodiment includes a trackside intelligent control device 21, a second ultraviolet light source 22, a second ultraviolet filter 23, and a second camera 24.

[0034] The trackside intelligent control device 21 is connected to the remote platform 3 via wireless communication, and communicates with the second ultraviolet light source 22 and the second camera 24 via wired or wireless data communication.

[0035] The trackside intelligent control device 21 is fixedly installed at a designated trackside position on a designated train track; the second ultraviolet light source 22 and the second camera 24 are respectively fixedly installed at two designated bottom positions on the designated train track; the linear scan direction of the second camera 24 is perpendicular to the train direction of the designated train track; the second ultraviolet filter 23 is fixedly installed at the front end of the lens of the second camera 24.

[0036] The second ultraviolet filter 23 is used to transmit light in a preset second ultraviolet band and to filter light in any band other than the second ultraviolet band.

[0037] The second ultraviolet light band in this embodiment of the invention is a pre-set light wave band range, which can be customized based on actual application requirements.

[0038] The second camera 24 is a line scan camera.

[0039] The dynamic acquisition equipment group 2 is used to acquire data from the ultraviolet light filtered full view of the undercarriage of the moving train passing through the current track position to obtain the corresponding first full view of the undercarriage; and based on the dynamic brightness threshold method, it performs oil leakage fault detection based on the first full view of the undercarriage to obtain the corresponding dynamic detection report and send it to the remote platform 3.

[0040] In another specific implementation of this invention, the dynamic acquisition device group 2 is specifically used to acquire data from the ultraviolet-filtered full-view image of the undercarriage of a moving train passing through the current track position to obtain the corresponding first full-view image of the undercarriage: The trackside intelligent control device 21 identifies the time period of each train passing through the current track position based on the built-in train passage sensing module; at the beginning of the current train passage time, it turns on the second ultraviolet light source 22 for continuous illumination and starts the second camera 24 to capture filtered images; at the end of the current train passage time, it turns off the second ultraviolet light source 22 and stops the shooting operation of the second camera 24, and obtains the corresponding first line array image sequence from the second camera 24; and according to the line array image stitching principle, it stitches the undercarriage length image according to the first line array image sequence and uses the stitched undercarriage length image as the corresponding first undercarriage full image.

[0041] Here, the principle of linear array stitching mentioned in the embodiments of the present invention is as follows: all linear arrays in the linear array sequence are stitched together in chronological order with time as the vertical stitching direction to obtain a vertical long image.

[0042] The first line array image sequence in this embodiment of the invention is composed of multiple first line array images arranged in chronological order, with all first line array images having the same image height and image width. The image width of the first complete vehicle underside image is the same as the image width of the first line array images, and the image height of the first complete vehicle underside image is the sum of the image heights of all the first line array images.

[0043] It should be noted that the train passage sensing module built into the trackside intelligent control device 21 can make real-time judgments on whether there is a train passing at the current moment. There are various technical implementation methods, such as judgment schemes based on magnetic steel sensors, judgment schemes based on visual sensors, and judgment schemes based on infrared sensors. These are all common and publicly available technical solutions, which will not be repeated here.

[0044] It should also be noted that when the train passage perception module of the trackside intelligent control device 21 identifies the train passage period for each time it passes the current track position, the specific implementation method is as follows: it makes a real-time judgment on whether a train has passed at the previous moment and at the current moment; if no train passed at the previous moment and a train has passed at the current moment, then the current moment is taken as the start moment of the current train passage period; if a train passed at the previous moment and no train has passed at the current moment, then the current moment is taken as the end moment of the current train passage period.

[0045] It should also be noted that when the train passage sensing module of the trackside intelligent control device 21 confirms the start time of each train passage period, it will also send a passage start signal to the trackside intelligent control device 21 simultaneously. After receiving the passage start signal, the trackside intelligent control device 21 will turn on the second ultraviolet light source 22 for continuous illumination and start the second camera 24 to capture filtered images.

[0046] It should also be noted that when the train passage sensing module of the trackside intelligent control device 21 confirms the end time of each train passage period, it will also send a passage end signal to the trackside intelligent control device 21 simultaneously. After receiving the passage end signal, the trackside intelligent control device 21 will turn off the second ultraviolet light source 22, stop the shooting operation of the second camera 24, and obtain the corresponding first linear array sequence from the second camera 24.

[0047] It should also be noted that after the second camera 24 starts shooting, it will perform high-frequency line array shooting at a predetermined line array shooting frequency until the shooting is stopped; after stopping, a corresponding first line array sequence will be generated.

[0048] In another specific implementation of this invention, the dynamic acquisition device group 2 is specifically used to send the corresponding dynamic detection report obtained from the oil leak fault detection based on the first full-view image of the vehicle underside using the dynamic brightness threshold method to the remote platform 3: Step E1: The trackside intelligent control device 21 identifies the brightness of each pixel in the full image of the first vehicle underside to obtain the corresponding pixel brightness; and uses the average value of all pixel brightness as the corresponding reference brightness; and uses the product of a preset second multiple and the reference brightness as the corresponding current brightness threshold, where the second multiple is a positive integer greater than 2; and records all pixels with a brightness not less than the current brightness threshold as fluorescent points; and counts the total number of fluorescent points to obtain the total number N2; and identifies whether the total number N2 is zero. Step E2: If N2=0, then set the corresponding vehicle under-body detection box set to empty; Step E3: If N2 > 0, cluster all fluorescent points in the first vehicle underside image to obtain one or more corresponding second point sets; identify whether the total number of fluorescent points in each second point set is less than a preset second total threshold; if so, delete the current point set; identify whether the total number of points in the remaining second point sets is zero; if so, set the corresponding vehicle underside detection box set to empty; if not, form a corresponding second detection box by the image coordinates of the four boundary points of each remaining second point set, and form the corresponding vehicle underside detection box set by all the obtained second detection boxes. Here, in this embodiment of the invention, at least one of the point distances between any point in the second point set and all other points in the current point set is less than a preset second point distance threshold. The second total threshold and the second point spacing threshold in this embodiment of the invention are two preset threshold parameters that can be customized based on actual application requirements; Step E4: The track marker and track installation position of the current equipment group are taken as the corresponding first track marker and first track position; the train passage time corresponding to the current first car bottom full map is taken as the corresponding first detection time period; and the corresponding dynamic detection report composed of the first track marker, first track position, first detection time period and car bottom detection frame set is sent to the remote platform 3.

[0049] (III) Remote Platform 3: The remote platform 3 in this embodiment of the invention is used to store multiple train inspection databases.

[0050] Here, in this embodiment of the invention, the train inspection database corresponds one-to-one with a train; the train inspection database is used to store all static inspection reports and dynamic inspection reports of the corresponding train.

[0051] The remote platform 3 is also used to store each received static or dynamic inspection report into the corresponding train inspection database.

[0052] Remote platform 3 is also used to periodically upgrade and train the visual fault recognition model and synchronize the latest trained model to all static acquisition device groups 1.

[0053] In another specific implementation of this invention, the remote platform 3 is specifically used to store each received static or dynamic inspection report into the corresponding train inspection database; Step F1: Upon receiving the static inspection report, extract the corresponding first train identifier from the current static inspection report; and store the current static inspection report in the train inspection database corresponding to the current first train identifier. Step F2: Upon receiving the dynamic detection report, extract the corresponding first track identifier, first track position, and first detection period from the current dynamic detection report; and through the preset nationwide train dispatch query interface, query the unique identifier of the train that travels along the track corresponding to the first track identifier and passes through the first track position within the first detection period, and use the query result as the corresponding current train identifier; and store the current dynamic detection report in the train detection database corresponding to the current train identifier.

[0054] Here, the network-wide train dispatch query interface in this embodiment of the invention is a pre-set processing interface that can be used to query train identifiers based on comprehensive information such as track, train location, and time.

[0055] In another specific implementation of this invention, the remote platform 3 is specifically used to periodically upgrade and train the visual fault recognition model and synchronize the latest trained model to all static acquisition device groups 1: On the platform side, large-scale big data collection is carried out on the undercarriage inspection images of all operating models on the entire network under various lighting conditions to obtain the current raw image set; and a corresponding model training dataset is constructed based on the raw image set; and a visual fault recognition model is trained once based on the model training dataset; and at the end of this round of training, the latest visual fault recognition model is synchronized to the undercarriage inspection robot 11 of all static acquisition equipment group 1.

[0056] Here, the original image set of this embodiment of the invention includes multiple undercarriage inspection images; each undercarriage inspection image is a visible light image. In the original image set, some undercarriage inspection images correspond to undercarriage areas with no loose bolts or damaged components, while others correspond to undercarriage inspection images with one or more loose bolts and / or damaged components. The undercarriage inspection images corresponding to undercarriage areas without loose bolts or damaged components do not have detection box labels; the undercarriage inspection images corresponding to undercarriage areas with loose bolts or damaged components have one or more corresponding detection box labels, with each detection box label corresponding one-to-one with the location of loose bolts or damaged components in the current undercarriage area.

[0057] This invention provides a processing system for inspecting the undercarriage of static and dynamic rail trains. As described above, the system includes: a static acquisition equipment group, a dynamic acquisition equipment group, and a remote platform. The system has at least the following technical effects or advantages: 1) The static acquisition equipment group uses an undercarriage robot to collect visible light and ultraviolet light filtered images of all designated component locations on the undercarriage of a static train parked on the tracks inside the depot. Based on a dynamic brightness threshold method, it performs oil leak detection using the ultraviolet image set, and based on a visual fault recognition model, it detects bolt loosening and component damage using the visible light image set. This not only improves the detection accuracy of lubricating oil leak areas but also further reduces the maintenance cycle of a single train, lowers inspection costs, improves inspection efficiency, and increases the maintenance frequency of each train; 2) The dynamic acquisition equipment group uses a line scan camera to collect data from the ultraviolet light filtered undercarriage of a moving train and performs oil leak detection based on the dynamic brightness threshold method. This not only compensates for the shortcomings of conventional solutions but also significantly increases the oil leak detection frequency of each train.

[0058] The steps of the system methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, software modules executed by a processor, or a combination of both. The software modules can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art.

[0059] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A processing system for detecting the underframe of a dynamic and static rail vehicle, characterized in that it comprises: The system includes: a static data acquisition device group, a dynamic data acquisition device group, and a remote platform; The remote platform is connected to the static acquisition equipment group and the dynamic acquisition equipment group respectively; the lubricating oil used in all undercarriage components of any train detected by the static acquisition equipment group and the dynamic acquisition equipment group contains fluorescent agent, and the lubricating oil is delivered to each component through a closed hydraulic pipeline; The static data acquisition equipment group is used to collect visible light and ultraviolet light filtered images of all designated component locations on the undercarriage of a static train parked on the tracks inside the garage through an undercarriage robot inspection, obtaining corresponding visible light image sets and ultraviolet light image sets; and to perform oil leakage fault detection based on the ultraviolet image sets using a dynamic brightness threshold method to obtain a corresponding first detection report; and to perform bolt loosening and component damage fault detection based on the visible light image sets using a visual fault recognition model to obtain a corresponding second detection report; and to generate a corresponding static detection report based on the current train's basic information and the first and second detection reports, and send it to the remote platform. The linear array camera of the dynamic acquisition device group is fixedly installed at the bottom of the designated track. The dynamic acquisition device group is used to acquire data from the ultraviolet filtered full-body image of the undercarriage of a moving train passing through the current track position to obtain the corresponding first full-body image of the undercarriage. Based on the dynamic brightness threshold method, the group performs oil leakage fault detection based on the first full-body image of the undercarriage and obtains the corresponding dynamic detection report, which is then sent to the remote platform. The remote platform is used to store multiple train inspection databases; each train inspection database corresponds one-to-one with a train; the train inspection database is used to store all the static inspection reports and dynamic inspection reports of the corresponding train; The remote platform is also used to store each received static or dynamic test report into the corresponding train test database; The remote platform is also used to periodically upgrade and train the visual fault recognition model and synchronize the latest trained model to all the static acquisition device groups.

2. The processing system for undercarriage inspection of dynamic and static rail trains according to claim 1, characterized in that, The static data acquisition equipment group includes an under-vehicle inspection robot, a six-degree-of-freedom robotic arm, a light source switching assembly, a filter switching assembly, and a first camera. The light source switching assembly has two built-in ambient light sources for the first camera: a first visible light source and a first ultraviolet light source. The filter switching assembly has two built-in lens filters for the first camera: a first visible filter and a first ultraviolet filter. The first visible filter transmits light within a preset visible light band and filters light of any band outside the visible light band. The first ultraviolet filter transmits light within a preset first ultraviolet light band and filters light of any band outside the first ultraviolet light band. The first camera is an area array camera. The under-vehicle inspection robot is connected to the remote platform via wireless communication and communicates with the six-degree-of-freedom robotic arm, the light source switching component, the filter switching component, and the first camera via wired or wireless data communication. The six-degree-of-freedom robotic arm is mounted on the under-vehicle inspection robot and can rotate at any angle under the scheduling and control of the under-vehicle inspection robot. The light source switching component, the filter switching component, and the first camera are installed at three fixed positions at the end of the six-degree-of-freedom robotic arm. The undercarriage inspection robot has a built-in visual fault recognition model; the visual fault recognition model is based on a type of visual target recognition model; the visual target recognition model includes the YOLO series model; the visual fault recognition model is used to perform target detection and classification recognition processing on the input image and output the corresponding set of target detection boxes. When the target detection box set is not empty, it consists of one or more target detection boxes; the target detection box includes the coordinates of the center point of the detection box, the height of the detection box, the width of the detection box, and the target type; the target type includes loose bolts and broken parts; The dynamic acquisition equipment group includes a trackside intelligent control device, a second ultraviolet light source, a second ultraviolet filter, and a second camera; the second ultraviolet filter is used to transmit light in a preset second ultraviolet light band and filter light in any band other than the second ultraviolet light band; the second camera is a line scan camera. The trackside intelligent control device is connected to the remote platform via wireless communication and communicates with the second ultraviolet light source and the second camera via wired or wireless data communication respectively; the trackside intelligent control device is fixedly installed at a designated trackside location on the designated train track; The second ultraviolet light source and the second camera are respectively fixedly installed at two designated bottom positions of the designated trolley track; the linear scan direction of the second camera is perpendicular to the trolley track's travel direction; the second ultraviolet filter is fixedly installed at the front end of the second camera's lens.

3. The processing system for undercarriage inspection of dynamic and static rail trains according to claim 2, characterized in that, The static acquisition equipment group is specifically used to acquire visible light and ultraviolet light filtered images of all designated component locations on the undercarriage of a static train parked on the tracks inside the garage through an undercarriage robot inspection, to obtain corresponding visible light image sets and ultraviolet light image sets: When the undercarriage inspection robot is dispatched by the operator to the initial inspection point position of any static train, it pauses and receives the first train identifier and the first train model input by the operator; and obtains the corresponding first inspection point map by querying the built-in undercarriage inspection point map library based on the first train model. And based on the first inspection point location map, the corresponding first inspection path is obtained by planning the inspection path. And start the current inspection process based on the first inspection path; During the current inspection, the built-in lidar or visual SLAM navigation system will perform motion navigation based on the first inspection path. And during the current inspection process, pause at the trajectory point position corresponding to each first inspection point on the first inspection point map; Each time the trajectory is paused, the six-degree-of-freedom robotic arm is first dispatched to send the first camera to the shooting position corresponding to the current vehicle under-point. Then, the light source switching component is dispatched to turn on the first visible light source, and the filter switching component is dispatched to switch the first visible filter for the first camera. The first camera is then dispatched to capture the filtered image to obtain the corresponding first visible light image. The light source switching component is then dispatched to turn off the first visible light source and turn on the first ultraviolet light source, and the filter switching component is dispatched to switch the first ultraviolet filter. The first camera is then dispatched to capture the filtered image to obtain the corresponding first ultraviolet light image. The light source switching component is then dispatched to turn off the first ultraviolet light source. The first point data is composed of the first point marker corresponding to the current inspection point and the first visible light image, and the second point data is composed of the current first point marker and the first ultraviolet light image. The current inspection process ends when the end point of the current inspection path is reached. At the end of this inspection process, the visible light image set is composed of all the data from the first point, and the ultraviolet light image set is composed of all the data from the second point. The vehicle undercarriage inspection point map library includes multiple first point map records; each first point map record includes a first vehicle model and a first inspection point map; each first inspection point map includes multiple first inspection points; each first inspection point corresponds to a designated component location on the undercarriage of the current vehicle model; each first inspection point includes a first point identifier and a first point relative position; the first point relative position includes a first lateral displacement and a first longitudinal displacement; the first lateral displacement and the first longitudinal displacement are the lateral and longitudinal relative displacement distances between the previous inspection point and the initial inspection point, respectively.

4. The processing system for undercarriage inspection of dynamic and static rail trains according to claim 3, characterized in that, The static acquisition device group is specifically used when the first detection report is obtained by detecting oil leak faults based on the ultraviolet light image set using the dynamic brightness threshold method: The undercarriage inspection robot uses each of the first ultraviolet images in the ultraviolet image set as the current ultraviolet image; The brightness of each pixel in the current ultraviolet light image is identified to obtain the corresponding pixel brightness. The average brightness of all the pixels is then used as the corresponding reference brightness. The product of the preset first multiple and the reference brightness is used as the corresponding current brightness threshold, where the first multiple is a positive integer greater than 2; All pixels whose brightness is not less than the current brightness threshold are recorded as fluorescent points; the total number of fluorescent points is counted to obtain a total number N1; and it is determined whether the total number N1 is zero. If N1=0, then set the corresponding first point detection box set to empty; If N1 > 0, then all the fluorescence points in the current ultraviolet image are clustered to obtain one or more corresponding first point sets; and it is identified whether the total number of fluorescence points in each first point set is less than a preset first total threshold. If so, the current point set is deleted; and it is identified whether the total number of points in the remaining first point sets is zero. If so, the corresponding first point detection box set is set to empty; if not, a corresponding first detection box is formed by the image coordinates of the four boundary points of the remaining first point sets, and the corresponding first point detection box set is formed by all the obtained first detection boxes; wherein, at least one of the point distances between any point in the first point set and all other points in the current point set is less than a preset first point distance threshold. The current ultraviolet light image and its corresponding first point identifier and the first point detection box set constitute the corresponding first point detection record; The first detection report is composed of all the detection records of the first point obtained.

5. The processing system for undercarriage inspection of dynamic and static track trains according to claim 3, characterized in that, The static acquisition equipment group is specifically used when the second detection report is obtained by detecting bolt loosening and component damage based on the visible light image set using the visual fault recognition model: The undercarriage inspection robot uses each of the first visible light images in the visible light image set as the current visible light image; The current visible light image is then used as the current model input image and fed into the visual fault recognition model for target detection and classification to obtain the corresponding set of target detection boxes. And the current set of target detection boxes is used as the corresponding set of second point detection boxes; The current visible light image and its corresponding first point identifier and second point detection box set constitute the corresponding second point detection record; The second detection report is composed of all the detection records of the second points.

6. The processing system for undercarriage inspection of dynamic and static track trains according to claim 3, characterized in that, The static data acquisition device group is specifically used when the corresponding static inspection report is generated based on the current train's basic information and the first and second inspection reports and sent to the remote platform: The vehicle undercarriage inspection robot consists of the start and end times of the current inspection process, which together form the first inspection period. The static inspection report, consisting of the first inspection period, the first train identifier, the first train model, the first inspection point map, the first inspection report, and the second inspection report corresponding to the current inspection process, is sent to the remote platform.

7. The processing system for undercarriage inspection of dynamic and static rail trains according to claim 2, characterized in that, The dynamic acquisition equipment group is specifically used to acquire data from the ultraviolet-filtered full-body image of the undercarriage of a moving train passing through the current track position to obtain the corresponding first full-body image of the undercarriage: The trackside intelligent control device identifies the time period of each train passing through the current track position based on the built-in train passage sensing module; at the beginning of the current train passage time, it turns on the second ultraviolet light source for continuous illumination and starts the second camera to capture filtered images; at the end of the current train passage time, it turns off the second ultraviolet light source and stops the second camera's shooting operation, and obtains the corresponding first line array image sequence from the second camera; and according to the line array image stitching principle, it stitches the undercarriage length image based on the first line array image sequence and uses the stitched undercarriage length image as the corresponding first undercarriage full image; The first linear array sequence is composed of multiple first linear arrays arranged in chronological order, with all first linear arrays having the same image height and width. The linear array stitching principle is to stitch all the linear arrays in the sequence in chronological order with time as the vertical stitching direction to obtain a vertical long image. The image width of the first full-view image of the vehicle's underside is the same as the image width of the first linear array, and the image height of the first full-view image of the vehicle's underside is the sum of the image heights of all the first linear arrays.

8. The processing system for undercarriage inspection of dynamic and static rail trains according to claim 2, characterized in that, The dynamic acquisition device group is specifically used when the corresponding dynamic detection report obtained from the oil leak fault detection based on the first full-view image of the vehicle underside using the dynamic brightness threshold method is sent to the remote platform: The trackside intelligent control device identifies the brightness of each pixel in the full image of the first vehicle underside to obtain the corresponding pixel brightness; and uses the average value of all the pixel brightness as the corresponding reference brightness; and uses the product of a preset second multiple and the reference brightness as the corresponding current brightness threshold, where the second multiple is a positive integer greater than 2. All pixels whose brightness is not less than the current brightness threshold are recorded as fluorescent points; the total number of fluorescent points is counted to obtain a total number N2; and it is determined whether the total number N2 is zero. If N2=0, then set the corresponding undercarriage detection box set to empty; If N2 > 0, then all the fluorescent points in the first vehicle underside full image are clustered to obtain one or more corresponding second point sets; and it is identified whether the total number of fluorescent points in each second point set is less than a preset second total threshold. If so, the current point set is deleted; and it is identified whether the total number of points in the remaining second point sets is zero. If so, the corresponding vehicle underside detection box set is set to empty; if not, a corresponding second detection box is formed by the image coordinates of the four boundary points of each remaining second point set, and the corresponding vehicle underside detection box set is formed by all the obtained second detection boxes; wherein, at least one of the point distances between any point in the second point set and all other points in the current point set is less than a preset second point distance threshold. The current equipment group's corresponding track marker and track installation position are used as the corresponding first track marker and first track position; the current train passage time corresponding to the first undercarriage full map is used as the corresponding first detection time period; and the dynamic detection report composed of the first track marker, the first track position, the first detection time period, and the undercarriage detection frame set is sent to the remote platform.

9. The processing system for undercarriage inspection of dynamic and static track trains according to claim 1, characterized in that, The remote platform is specifically used when storing each received static or dynamic test report into the corresponding train test database; Upon receiving the static inspection report, the corresponding first train identifier is extracted from the current static inspection report; And store the current static inspection report into the train inspection database corresponding to the current first train identifier; Upon receiving the dynamic detection report, extract the corresponding first track identifier, first track position, and first detection period from the current dynamic detection report; And through the preset network-wide train dispatch query interface, the unique identifier of the train that travels along the track corresponding to the first track identifier and passes through the first track position during the first detection period is queried and the query result is used as the corresponding current train identifier. The current dynamic detection report is then stored in the train detection database corresponding to the current train identifier.

10. The processing system for undercarriage inspection of dynamic and static rail trains according to claim 1, characterized in that, The remote platform is specifically used when the visual fault recognition model is periodically upgraded and trained, and the latest trained model is synchronized to all the static acquisition device groups: On the platform side, large-scale big data collection is performed on under-vehicle inspection images of all operating models on the entire network under various lighting conditions to obtain the current raw image set; and a corresponding model training dataset is constructed based on the raw image set. The visual fault recognition model is then trained once using the model training dataset. At the end of this training round, the latest visual fault recognition model will be synchronized to the undercarriage inspection robots of all the static acquisition device groups; The original image set includes multiple undercarriage inspection images; each undercarriage inspection image is a visible light image; in the original image set, some undercarriage inspection images correspond to areas with no loose bolts or damaged components, while others correspond to areas with one or more loose bolts and / or damaged components; the undercarriage inspection images corresponding to areas without loose bolts or damaged components do not have detection box labels; the undercarriage inspection images corresponding to areas with loose bolts or damaged components have one or more corresponding detection box labels, and the detection box labels correspond one-to-one with the locations of loose bolts or damaged components in the current undercarriage area.

Images