Maglev train current collector detection method and device, computer device and storage medium

Abstract

The application relates to a maglev train current collector detection method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a plurality of original point cloud sets corresponding to a plurality of to-be-detected structures in a maglev train current collector and a plurality of target structure images; using each spatially transformed point cloud model to segment the corresponding original point cloud set to obtain a corresponding optimized point cloud set; finding two edge point cloud sets in each optimized point cloud set along the vertical direction of the running direction of the maglev train, and then obtaining a first straight line corresponding to each optimized point cloud set; finding the edge pixel positions in each target structure image along the vertical direction of the running direction of the maglev train, and then obtaining a second straight line; according to the slope of the second straight line and a corresponding threshold value, a slope threshold range is obtained; and according to the size relationship between the slope of the first straight line and the slope threshold range, whether the corresponding to-be-detected structure of the current collector is abnormal is detected. The method can improve the current collector detection efficiency, has high detection precision and stable detection effect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of maglev train current collector detection, in particular to a maglev train current collector detection method and device, computer equipment and a storage medium. BACKGROUND

[0002] As a new generation of transportation tool, the medium-low speed maglev train has many advantages such as fast speed, strong adaptability to terrain, flexible line selection, safety and environmental protection. The current collection system of the medium-low speed maglev train is composed of a current collector installed below the suspension frame and a contact rail in the waist part of the track beam. The current collector slide plate transmits and receives electric energy through sliding contact with the contact rail. Changsha maglev line and Qingyuan maglev both adopt three-rail catenary current collection type. According to the relative position of the current collector and the contact rail, the three-rail catenary is divided into upper contact type, lower contact type and side contact type. The current collector with side current collection is relatively simple, and its installation structure is compact and does not exceed the specified limit of the vehicle. The current collector adopts a hinged structure and is installed on the mounting bracket on the side of the support arm. Through mechanical sliding contact with the power rail, high voltage is sent to the maglev train to realize power supply to the medium-low speed maglev train. Therefore, the integrity of the current collector will directly affect the power supply system of the maglev train and further affect the running safety of the train. The detection of the integrity indicators such as damage, deformation and displacement of the current collector and the detection of the thickness of the carbon slide plate are of great significance to improve the maintenance efficiency and reduce the fault rate of the train online.

[0003] The three-dimensional detection technology based on point cloud registration has been widely applied in the field of industrial manufacturing. Taking the application in the part surface quality detection as an example, the three-dimensional detection technology realizes the comparison and evaluation by subtracting the two point clouds after registering the scanning point cloud of the target part with the corresponding standard model.

[0004] However, due to the influence of the operation of the maglev train, the installation position of the collection device and the complex hinged hollow structure of the current collector, it is difficult to obtain complete point cloud data of the current collector, and only a certain range of point cloud of the current collector can be obtained, thereby affecting the accuracy and stability of the point cloud registration. SUMMARY

[0005] Therefore, it is necessary to provide a maglev train current collector detection method, device, computer equipment and storage medium capable of improving the detection accuracy of the maglev train current collector in view of the above technical problems.

[0006] A maglev train current collector detection method, the method comprising:

[0007] Obtaining a plurality of original point cloud sets corresponding to a plurality of to-be-detected structures in the maglev train current collector and a plurality of target structure images;

[0008] Each original point cloud set is registered with the point cloud model of the corresponding structure to be detected to obtain the corresponding spatially transformed point cloud model. The corresponding original point cloud set is then segmented using each spatially transformed point cloud model to obtain the corresponding optimized point cloud set.

[0009] Find two edge point cloud sets along the vertical direction of the maglev train in each optimized point cloud set. Based on the edge point cloud coordinates in the two edge point cloud sets, obtain the first straight line corresponding to each optimized point cloud set. Find the edge pixel position along the vertical direction of the maglev train in each target structure image. Based on the edge pixel position, obtain the second straight line corresponding to the target structure image.

[0010] Based on the slope of the second line corresponding to the second line and the corresponding threshold, the slope threshold range is obtained. Based on the relationship between the slope of the first line corresponding to each optimized point cloud set and the slope threshold range, the abnormality of the multiple structures to be detected in the current receiver is detected.

[0011] In one embodiment, the method further includes: acquiring a current receiver point cloud corresponding to the current receiver of the maglev train; the current receiver of the maglev train includes multiple target structures to be detected; and segmenting the current receiver point cloud using a pre-set dividing line to obtain multiple original point cloud sets that each contain the multiple target structures to be detected.

[0012] In one embodiment, the method further includes: constructing a spatial rectangular coordinate system; the X-axis of the spatial rectangular coordinate system is perpendicular to the direction of operation of the maglev train, and the Y-axis of the spatial rectangular coordinate system is parallel to the direction of operation of the maglev train; acquiring the edge point cloud of the current receiver point cloud in the X-axis direction, and obtaining two dividing lines based on the edge point cloud coordinates in the two edge point cloud sets, the physical parameters of each structure of the current receiver, and the spatial topological relationship; using the two dividing lines to segment the current receiver point cloud to obtain a first original point cloud set, a second original point cloud set, and a third original point cloud set, each containing the plurality of structures to be detected.

[0013] In one embodiment, the method further includes: calculating the overlap degree between each original point cloud set and the corresponding registered point cloud set; if the overlap degree is not within a preset overlap degree threshold range, then outputting the detection result for the corresponding structure to be detected; the detection result is abnormal; if the overlap degree is within the preset overlap degree threshold range, calculating the mean of the overlapping region between each original point cloud set and the corresponding registered point cloud set; if the mean of the overlapping region is not within a preset mean threshold range, then outputting the detection result for the corresponding structure to be detected; the detection result is abnormal; if the mean of the overlapping region is within the preset mean threshold range, calculating the variance of the overlapping region between each original point cloud set and the corresponding registered point cloud set; if the variance of the overlapping region is not within a preset variance threshold range, then outputting the detection result for the corresponding structure to be detected; the detection result is abnormal.

[0014] In one embodiment, the method further includes: segmenting the corresponding original point cloud set according to the boundary information of each spatially transformed point cloud model, extracting overlapping region point clouds, clustering the overlapping region point clouds, removing point cloud clusters with a small number of neighboring point clouds, and obtaining the corresponding optimized point cloud set.

[0015] In one embodiment, the method further includes: acquiring a test image of the maglev train current collector; and processing the test image using a contour matching method to obtain multiple target structure images corresponding to the multiple test structures.

[0016] In one embodiment, the method further includes: acquiring the train speed and the fluctuation amount perpendicular to the train's direction of travel at each moment; determining whether the image to be detected has motion blur based on the relationship between the fluctuation amount and the image precision within the fluctuation time; when motion blur exists in the image, adjusting the exposure time of the current camera according to the train's current speed, and updating the image to be detected based on the adjusted exposure time.

[0017] A current collector detection device for maglev trains, the device comprising:

[0018] The data acquisition module is used to acquire multiple sets of original point clouds and multiple target structure images corresponding to multiple structures to be detected in the current collector of the maglev train;

[0019] The point cloud optimization module is used to perform point cloud registration between each original point cloud set and the point cloud model of the corresponding structure to be detected, so as to obtain the corresponding spatially transformed point cloud model. The corresponding original point cloud set is segmented using each spatially transformed point cloud model to obtain the corresponding optimized point cloud set.

[0020] The straight line fitting module is used to find two edge point cloud sets along the vertical direction of the maglev train in each optimized point cloud set, obtain the first straight line corresponding to each optimized point cloud set based on the edge point cloud coordinates in the two edge point cloud sets, find the edge pixel position along the vertical direction of the maglev train in each target structure image, and obtain the second straight line corresponding to the target structure image based on the edge pixel position.

[0021] The structure detection module is used to obtain a slope threshold range based on the slope of the second straight line corresponding to the second straight line and the corresponding threshold, and to detect whether the multiple structures to be detected of the current receiver are abnormal based on the relationship between the slope of the first straight line corresponding to each optimized point cloud set and the slope threshold range.

[0022] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program performing the following steps:

[0023] Acquire multiple original point cloud sets and multiple target structure images corresponding to multiple structures to be detected in the current collector of a maglev train;

[0024] Each original point cloud set is registered with the point cloud model of the corresponding structure to be detected to obtain the corresponding spatially transformed point cloud model. The corresponding original point cloud set is then segmented using each spatially transformed point cloud model to obtain the corresponding optimized point cloud set.

[0025] Find two edge point cloud sets along the vertical direction of the maglev train in each optimized point cloud set. Based on the edge point cloud coordinates in the two edge point cloud sets, obtain the first straight line corresponding to each optimized point cloud set. Find the edge pixel position along the vertical direction of the maglev train in each target structure image. Based on the edge pixel position, obtain the second straight line corresponding to the target structure image.

[0026] Based on the slope of the second line corresponding to the second line and the corresponding threshold, the slope threshold range is obtained. Based on the relationship between the slope of the first line corresponding to each optimized point cloud set and the slope threshold range, the abnormality of the multiple structures to be detected in the current receiver is detected.

[0027] A computer-readable storage medium having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0028] Acquire multiple original point cloud sets and multiple target structure images corresponding to multiple structures to be detected in the current collector of a maglev train;

[0029] Each original point cloud set is registered with the point cloud model of the corresponding structure to be detected to obtain the corresponding spatially transformed point cloud model. The corresponding original point cloud set is then segmented using each spatially transformed point cloud model to obtain the corresponding optimized point cloud set.

[0030] Find two edge point cloud sets along the vertical direction of the maglev train in each optimized point cloud set. Based on the edge point cloud coordinates in the two edge point cloud sets, obtain the first straight line corresponding to each optimized point cloud set. Find the edge pixel position along the vertical direction of the maglev train in each target structure image. Based on the edge pixel position, obtain the second straight line corresponding to the target structure image.

[0031] Based on the slope of the second line corresponding to the second line and the corresponding threshold, the slope threshold range is obtained. Based on the relationship between the slope of the first line corresponding to each optimized point cloud set and the slope threshold range, the abnormality of the multiple structures to be detected in the current receiver is detected.

[0032] The aforementioned method, apparatus, computer equipment, and storage medium for detecting current collectors in maglev trains register the original point clouds corresponding to multiple structures to be detected in the current collector using a point cloud registration method. Then, the corresponding original point cloud set is segmented using a spatially transformed point cloud model to remove interfering point clouds and obtain a more accurate optimized point cloud. Next, a first straight line fitting is performed based on the edge point cloud positions along the perpendicular direction of train operation in the optimized point cloud set. A second straight line fitting is performed based on the edge pixel positions along the perpendicular direction of train operation in multiple target structure images. This quantifies whether the structure to be detected has undergone significant movement during train operation. Furthermore, by acquiring the slope information of the straight lines in the two-dimensional image, the influence caused by incomplete acquisition of current collector point cloud data can be reduced. Finally, by comparing the slopes of the first and second straight lines and their corresponding thresholds, the abnormality of the structure to be detected in the current collector is detected. This invention significantly improves the efficiency of current collector detection, with high detection accuracy and stable detection results. It eliminates false detections and missed detections caused by manual inspection, thereby improving vehicle safety. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating a method for detecting the current collector of a maglev train in one embodiment.

[0034] Figure 2 This is a schematic diagram of the structure of a maglev train current collector detection device in one embodiment;

[0035] Figure 3 This is a flowchart illustrating a specific embodiment of a method for detecting the current collector of a maglev train.

[0036] Figure 4 This is a schematic diagram of the current collector of a maglev train in another embodiment;

[0037] Figure 5 This is a flowchart illustrating a method for detecting the thickness of carbon sliding plates on a maglev train in one embodiment.

[0038] Figure 6 This is a flowchart illustrating the contour matching method in one embodiment, wherein (a) is a schematic diagram of a skateboard side image template, (b) is a real-shot skateboard side image, (c) is a schematic diagram of the skateboard side contour, (d) is a real-shot skateboard side edge image, and (e) is the contour matching result;

[0039] Figure 7 This is a schematic diagram of the sliding matching range in one embodiment;

[0040] Figure 8 This is a schematic diagram of the process for extracting the thickness of a carbon skateboard in one embodiment, wherein (a) is a schematic diagram of removing edge points outside the skateboard position frame, (b) is a schematic diagram of removing edge points in the upper half, (c) is a schematic diagram of Hough transform line detection, and (d) is a schematic diagram of determining the upper and lower surfaces of the skateboard body.

[0041] Figure 9 This is a structural block diagram of a maglev train current collector detection device in one embodiment;

[0042] Figure 10 This is an internal structural diagram of a computer device in one embodiment; Attached image description:

[0044] 1. Maglev train; 2. Speed ​​measuring device; 3. Start-stop device; 4. Train number acquisition device; 5. Suspension frame; 50. Current collector; 6. Data acquisition device; 501. Carbon slide plate; 502. Carbon slide plate base; 503. Support base; 504. Swing arm; 505. Torsion spring; 506. Support arm. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0046] like Figure 2 As shown, a schematic diagram of a current collector detection device for a maglev train is provided.

[0047] The system is equipped with an online / offline photoelectric switch and an integrated controller. In non-operating mode, the system is in sleep mode to reduce energy consumption. Start-stop device 3 triggers the system's operating signal when a vehicle passes the photoelectric switch, initiating detection. When the vehicle finishes detection and passes the photoelectric switch again, it triggers the system's sleep signal, returning the system to sleep mode. The online / offline photoelectric switch is a through-beam photoelectric sensor (OBE20M-R100-S2EP-IO-L), using a 680nm wavelength laser, capable of measuring a maximum distance of 30m, with a response time of 0.4ms, an operating temperature range of -40 to 60℃, and an IP69K protection rating.

[0048] When the train passes through the detection system, it operates in manual driving mode, resulting in a non-constant train speed. Due to this inconsistent speed, the camera cannot be directly triggered for scanning via pulses. Therefore, this invention introduces a speed measuring device 2. By acquiring the train speed, the camera exposure time is set to resolve the image blurring issue caused by rapid shooting. The speed measuring device 2 employs a customized narrow-wave lidar with a speed measuring range of 0 to 50 km / h, a speed measuring accuracy of 0.1 km / h, a response time of less than 30 ms, and an IP67 protection rating.

[0049] Since the installation location of the current collector on the train is known, and the accurate train speed is obtained through a laser velocimeter, the moment when the current collector passes through the field of view of the detection camera can be calculated relatively accurately. The acquisition device 6 uses 0.2 seconds before and after this moment as the detection window, acquiring 2D images of the current collector at a high speed of 60 frames per second. A 3D camera acquires point cloud data of the current collector. Image processing is performed during the intervals between acquisition windows, which can significantly reduce the system's false detection rate and reduce the system's requirements for high-speed real-time image processing. This allows for the detection of current collectors on maglev trains at different speeds.

[0050] In one embodiment, such as Figure 1 As shown, a method for detecting the current collector of a maglev train is provided, including the following steps:

[0051] Step 102: Obtain multiple sets of original point clouds and multiple target structure images corresponding to multiple structures to be detected in the current collector of the maglev train.

[0052] Traditional two-dimensional image detection methods for inspecting current receivers are limited by dimensionality, only allowing detection of the integrity of one plane of the receiver. Three-dimensional detection technology, however, can acquire more information about the current receiver, significantly improving train operation safety. Figure 4The schematic diagram of the current collector of the maglev train shown is illustrated. The point cloud coordinate system uses the direction perpendicular to the train's running direction as the X-axis, the direction of the maglev train's running direction as the Y-axis, and the upward direction perpendicular to the train's running direction as the positive Z-axis. The current collector includes a carbon sliding plate 501, a carbon sliding plate base 502, a support base 503, a swing arm 504, a torsion spring 505, and a support arm 506. During normal train operation, the carbon sliding plate 501 contacts the power supply rail, and a certain force is applied to it. The support base 503, swing arm 504, and torsion spring 505 will move to some extent due to this force and changes in the train's state, ensuring good contact between the carbon sliding plate and the power supply rail. When performing current collector integrity testing, multiple structures to be tested, including the carbon sliding plate 501, carbon sliding plate base 502, swing arm 504, and support arm 506 of the current collector, are involved. The steps for obtaining multiple original point cloud sets corresponding to multiple structures to be tested in the maglev train current collector are as follows:

[0053] First, the acquired point cloud data is preprocessed by filtering to remove noise points and background point clouds, retaining only the point cloud data set within a certain Z-value range. Secondly, find the maximum. Value and minimum The values ​​are used to obtain the edge point cloud sets in both X-axis directions, and the point cloud coordinates in the two edge point cloud sets are calculated. The average of the values ​​were obtained respectively. and By combining the fixed physical parameters and spatial topological relationships of each structure of the current receiver, two dividing lines are obtained respectively. and ,in, Indicates the first segmentation parameter. This represents the second segmentation parameter, specifically, Add half the length of the torsion spring 505 to the length of the support arm 506 in the X-axis direction. The length of the carbon slide plate 501 and carbon slide plate base 502 in the X-axis direction is increased by half the length of the support base 503 to prevent incomplete point clouds of each part after segmentation, thus enabling complete segmentation of the point clouds of each structure to be detected. and For sets Partitioning yields a set. , and , where set This represents the first original point cloud set containing carbon skateboard 501 and carbon skateboard base 502. This represents the second original point cloud set containing swing arm 504. This represents the third set of original point clouds containing arm 506.

[0054] Furthermore, the acquired target structure images corresponding to multiple structures to be detected include a first target structure image containing carbon skateboard 501 and carbon skateboard base 502, a second target structure image containing swing arm 504, and a third target structure image containing support arm 506. Acquiring and processing two-dimensional images allows for more accurate straight line fitting. Comparing point cloud detection results with image detection results reduces measurement errors and improves detection accuracy. When detecting target structure images, contour matching is first used to identify the shape of the structure to be detected, and then Hough transform is used for feature extraction.

[0055] Step 104: Perform point cloud registration between each original point cloud set and the point cloud model of the corresponding structure to be detected to obtain the corresponding spatially transformed point cloud model. Use each spatially transformed point cloud model to segment the corresponding original point cloud set to obtain the corresponding optimized point cloud set.

[0056] The point cloud model is a pre-recorded standard point cloud model, including the first point cloud model containing the carbon skateboard 501 and the carbon skateboard base 502. The second point cloud model including swing arm 504 and the third point cloud model including the support arm 506 According to the pre-entered point cloud models , and For each set , , Perform point cloud registration and calculate the transformation matrix. , and Each model is multiplied by its corresponding matrix to obtain the spatially transformed point cloud model. , and .

[0057] To confirm structural integrity from multiple perspectives, based on the spatially transformed point cloud model... , and Boundary information for the original point cloud set , and Partitioning yields a new set. , and Specifically, by enlarging the point cloud of the spatially transformed point cloud model and extracting the overlapping area point cloud with the corresponding original point cloud set, more interfering point clouds can be removed, resulting in a more accurate point cloud. It is obtained by cropping the original point cloud set based on the model, and cannot guarantee complete overlap with the model. , and Clustering is performed separately, and point cloud clusters with a small number of neighboring point clouds are removed to obtain an optimized point cloud set. In practical applications, this step can remove point clouds of protruding bolts, similar to those on a carbon skateboard base.

[0058] Step 106: Find two edge point cloud sets along the vertical direction of the maglev train in each optimized point cloud set. Based on the edge point cloud coordinates in the two edge point cloud sets, obtain the first straight line corresponding to each optimized point cloud set. Find the edge pixel position along the vertical direction of the maglev train in each target structure image. Based on the edge pixel position, obtain the second straight line corresponding to the target structure image.

[0059] After the camera is mounted and fixed, the carbon slide plate 501, carbon slide plate base 502, and support arm 506 can only move within a very small range. The swing arm 504 can be compressed around the torsion spring 505, but the slopes of the two straight lines of the swing arm 504 are similar. Therefore, for each optimized point cloud set, the X-axis direction is taken. The set of point clouds with the largest value The point cloud set with the smallest value is then processed using RANSAC (RANdom Sampling Consensus) algorithm. The set of point clouds with the largest value The point cloud set with the smallest value is sampled and a straight line is fitted.

[0060] During the movement of the maglev train, the data acquisition device takes pictures from bottom to top. Due to the limitations of the installation position and the complex hinged and hollow structure of the current receiver, there will be occlusion and point cloud holes. It can only collect point clouds of the current receiver within a certain range. Therefore, in order to ensure the accuracy of the straight line fitting based on the three-dimensional point cloud, image processing methods are used to process the two-dimensional image of the current receiver to obtain a more accurate straight line through the corresponding image pixel positions. The slope of the straight line measured by the two methods is compared and detected to determine whether the slope range is within the threshold requirement. If it is not within the range, it indicates that deformation may have occurred, and the program will issue a warning.

[0061] Step 108: Based on the slope of the second straight line and the corresponding threshold, obtain the slope threshold range. Based on the relationship between the slope of the first straight line corresponding to each optimized point cloud set and the slope threshold range, detect whether multiple structures to be detected in the current receiver are abnormal.

[0062] In the aforementioned method for detecting current collectors in maglev trains, point cloud registration is used to register the original point clouds corresponding to multiple structures to be detected in the current collector. The corresponding original point cloud set is then segmented using a spatially transformed point cloud model to remove interfering point clouds and obtain a more accurate optimized point cloud. Next, a first straight line fitting is performed based on the edge point cloud positions along the perpendicular direction of train operation in the optimized point cloud set. A second straight line fitting is performed based on the edge pixel positions along the perpendicular direction of train operation in multiple target structure images. This quantifies whether the structure to be detected has undergone significant movement during train operation. Furthermore, by acquiring the slope information of the straight lines in the two-dimensional image, the influence caused by incomplete acquisition of the current collector point cloud data can be reduced. Finally, by comparing the slopes of the first and second straight lines and their corresponding thresholds, the abnormality of the structure to be detected in the current collector is detected. This embodiment of the invention can significantly improve the detection efficiency of the current collector, with high detection accuracy and stable detection results. It eliminates false detections and missed detections caused by manual inspection, thereby improving vehicle safety.

[0063] When capturing two-dimensional images of the current collector, the speed and directional fluctuations of the maglev train can cause deviations in the measurement results. The camera exposure settings directly affect the image clarity and sharpness, making the monitoring of speed and fluctuation directions, as well as the handling of image motion blur, particularly crucial. First, the relationship between the fluctuation amount and image accuracy is used to determine whether the degree of motion blur will affect the image quality. Fluctuation amount, also called lateral displacement, is crucial. Medium- and low-speed maglev trains primarily rely on the self-resetting ability of levitation force for guidance. When the train passes through curves, centrifugal force causes the vehicle to move laterally, and the magnetic field lines within the levitation gap are twisted, forming a lateral electromagnetic force. The guiding force is the lateral component of the vertical levitation force. According to Appendix A of the "Design Specification for Medium- and Low-Speed ​​Maglev Transportation CJJ / T 262-2017," the dynamic lateral displacement of the maglev frame relative to the F-rail magnetic poles is ±14mm. Second, when image motion blur occurs, the exposure time is adjusted / set in real-time according to the train speed to reduce the degree of image motion blur. The specific steps are as follows:

[0064] First, establish a spatial rectangular coordinate system, with the direction of the maglev train's head as the positive X-axis. According to the right-hand rectangular coordinate system, the direction perpendicular to the train's direction of travel is the Y-axis, and the upward direction perpendicular to the train's direction of travel is the positive Z-axis. Obtain the train's speed along the X-axis. Fluctuation in the Y-axis direction Construct timestamps respectively With running speed and volatility The model uses the model timestamp as the standard to obtain the instantaneous YOZ fluctuation vector in the plane. A model of the instantaneous fluctuation vector changing over time is established, with the timestamp as the X-axis. Based on the instantaneous fluctuation vector... Calculate the Y-axis ripple: Fluctuation time: ,in, Indicates the timestamp at which the exposure began. The timestamp indicating the end of exposure is used to determine the relationship between fluctuation and accuracy, as shown in the following formula:

[0065]

[0066] in, The short-side field of view is defined as 1200, which represents the resolution of the short side of the field of view. When the fluctuation is less than the accuracy threshold, it does not affect the image quality. When the fluctuation is greater than the accuracy threshold, it indicates the presence of motion blur. In this embodiment of the invention, a 200W pixel camera with a resolution of 1600 is used. 1200, obtained by acquiring train speed With timestamp A model of speed variation over time is established, and the formula for calculating exposure time is as follows:

[0067]

[0068] in, For the longer side of the field of view, For the short side field of view, For the exposure time, the corresponding exposure time is set in real time according to the feedback train speed to ensure that the impact of image motion blur is minimized each time.

[0069] Furthermore, the sliding plate, as a device connecting the current collector and the contact rail conductor, directly affects the power source for the electric train's operation. The sliding plate directly connects the current collector and the contact rail, providing power to the maglev train from the contact rail whether stationary or sliding. During operation, the sliding plate is directly exposed to the natural environment and constantly experiences friction and impact with the contact rail. Routine maintenance requires checking the wear and tear of the sliding plate. The Changsha maglev train, which is already in mature operation, uses a metal-coated carbon sliding plate with a normal thickness range of 7.5mm-20mm. The thickness of the sliding plate is a crucial indicator of its remaining lifespan. Currently, the thickness of the sliding plate for medium- and low-speed maglev trains is mainly measured manually, by visual inspection or with a steel ruler. This method suffers from drawbacks such as poor maintenance environment, low measurement accuracy, high labor costs, and potential safety hazards associated with manual inspection. Regarding service life, the different conditions of Changsha maglev skateboards during use inside and outside the depot were analyzed. In sunny weather, the skateboard life is 6000km to 8000km, while in rainy weather, the skateboard wears out rapidly and the skateboard life is about 1000km. The method of this invention also includes a method for detecting the thickness of carbon skateboards.

[0070] The system uses two industrial-grade high-definition cameras to capture side images of the current collector slides on both sides of the train. The cameras have a resolution of 2 megapixels and a frame rate of up to 60 frames per second. Because the current collectors on the maglev train are positioned low, the cameras are installed near the lower railbed of the elevated track before entering the train maintenance workshop. When the train passes at a constant speed not exceeding 10 km / h, clear side images of the slides can be captured from below. Each camera is equipped with a strip LED light source for supplemental lighting to ensure a good lighting environment during shooting. The start and stop of the system's detection process are controlled by the train's offline photoelectric pair.

[0071] When the maglev train approaches the maintenance workshop at a constant speed not exceeding 10 km / h, the incoming photoelectric sensor first sends an incoming signal. At this time, the camera starts and the system begins detection. As the train passes the camera, the system continues to detect the thickness of all sliding plates on the train in sequence. The detection results are then saved and transmitted to the operation control center. When the train passes the offline photoelectric sensor, the offline photoelectric sensor sends an offline signal, and the system stops detection. The entire detection system can detect the thickness of all sliding plates on the train in a single pass.

[0072] A three-car medium-low speed maglev train has four sliding plates on each side, which pass through the field of view of the camera in sequence. Before thickness detection, it is necessary to identify whether there is a side image of the sliding plate in the current image and its location. Since the distance between the sliding plate and the camera is fixed, the shape and size of the side of the sliding plate in the image is almost unchanged. Therefore, this invention uses the contour matching method to identify the shape of the sliding plate.

[0073] In one embodiment, such as Figure 5 The diagram illustrates a flowchart of a method for detecting the thickness of a carbon skateboard on a maglev train. The specific steps of the carbon skateboard thickness detection method are as follows: Edge detection is performed based on the skateboard side image template and the image to be detected to obtain a skateboard side contour template; the skateboard side contour template is then used for sliding matching in the edge-detected image to be detected, and the matching value between the skateboard side contour template and the skateboard side contour in the image to be detected is calculated. When the matching value reaches a preset matching threshold, the matching result is output; based on the matching result, the upper and lower surface line segments of the skateboard body are extracted, and the average distance between the upper and lower surface line segments is calculated to obtain the skateboard thickness.

[0074] Specifically, such as Figure 6The diagram illustrates a flowchart of a contour matching method. (a) is a template of a skateboard side image, (b) is a real-shot skateboard side image, (c) is a schematic diagram of the skateboard side contour, (d) is a real-shot skateboard side edge image, and (e) is the contour matching result. First, a template of the skateboard side image is obtained. The Sobel operator is used for edge detection, and sparse points in the edge detection results are removed to obtain the contour of the skateboard side. Then, edge detection is also performed on the acquired skateboard side image, and sparse points in the results are removed. Finally, the skateboard side contour is slide-matched against the real-shot skateboard side edge image. The bounding box with the highest matching value is the output of the contour matching. The matching value is calculated using the following formula:

[0075] ,

[0076] During sliding matching, for each contour point in the skateboard's side profile, if there is also an edge point within a very small range of its corresponding point in the actual skateboard side edge image, then this point is considered a match. Before finally outputting the matching result, it is also necessary to check whether the highest matching value reaches the threshold T. If the highest matching value does not reach the threshold T, then the skateboard is considered not to exist in the image. Figure 7 As shown. In practice, T is usually set to 0.8, so that even if the quality of the captured image is occasionally poor, a matching result can still be output. In this embodiment, since the camera and contact rail are fixed, and the slide plate is always in contact with the contact rail, the vertical position of the slide plate side in the image always changes very little. Taking advantage of this, the sliding range can be greatly reduced during sliding matching, thereby effectively reducing the amount of computation and greatly improving the matching speed.

[0077] like Figure 8As shown, a flowchart for extracting the thickness of a carbon skateboard is provided. (a) illustrates the removal of edge points outside the skateboard's position frame, (b) illustrates the removal of edge points in the upper half, (c) illustrates Hough transform line detection, and (d) illustrates the determination of the upper and lower surfaces of the skateboard body. After obtaining the position frame of the skateboard side in the image through skateboard shape recognition, this result can greatly simplify the detection of the skateboard thickness. The specific steps include: First, removing edge points outside the skateboard's position frame from the actual skateboard side edge image. Then, removing edge points in the upper half of the skateboard position frame, as the upper half is the skateboard support frame and is unrelated to the skateboard body. Next, using Hough transform on the remaining edge points to find long straight lines. The angle of the candidate straight lines is very small during the transformation, so the transformed lines are nearly horizontal, meeting the requirements while effectively avoiding interference. Finally, the two lowest straight lines are selected as the upper and lower surfaces of the skateboard body. By restricting the obtained upper and lower surface lines of the skateboard body within the skateboard position frame, the line segments of the upper and lower surfaces of the skateboard body can be obtained. The average distance between these two line segments is calculated, and the camera calibration parameters are used to calculate the accurate value of the skateboard thickness.

[0078] In one embodiment, the step of extracting line segments from the upper and lower surfaces of the skateboard body based on the matching result includes: obtaining the vehicle body tilt angle based on the current suspension gap; and extracting the upper and lower surface line segments of the skateboard body based on the vehicle body tilt angle and the matching result when extracting the thickness of the carbon skateboard in the updated image to be detected.

[0079] In this embodiment, the carbon slide plate to be inspected should ideally be parallel to the camera chip plane. However, in actual operation during train maintenance, the carbon slide plate will have a certain angle, which will affect the actual thickness in high-precision inspections. A maglev train consists of three carriages. Each carriage has four suspension gap sensors installed on each suspension frame. Each sensor has four probes. The camera acquires the suspension gap at the nearest acquisition point at the same time as it begins to acquire images, and constructs a timestamp. and gap The model (Z-axis direction) calculates the vehicle body tilt angle based on the change in suspension gap on both sides of the vehicle body:

[0080]

[0081] in, express The difference in the floating distance between the two sides of the vehicle body at any given time. This indicates the physical distance between the opposing magnetic levitation sensors; when extracting thickness, the angle is added. Make corrections to obtain more accurate results.

[0082] In one specific embodiment, the lower limit of the carbon slide plate thickness is 7.5mm, and the system detection accuracy is 0.5mm. Therefore, 8mm is taken as the lower limit of the carbon slide plate thickness detection result. Statistical analysis of the carbon slide plate wear is performed to calculate the theoretical daily wear thickness. (mm), the carbon slide plate thickness detected by the method of this invention is (mm), based on the test results, early warning analysis and wear anomaly analysis are performed:

[0083] like If the value is ≤8, the system will alarm and prompt maintenance personnel to replace the carbon slide plate.

[0084] If 8 < ≤ (8+7) If the carbon slide plate is not in good condition, the system will issue a warning, prompting maintenance personnel to check its condition and prepare for replacement.

[0085] like > (8+7) If the carbon slide plate thickness is normal, no inspection is required.

[0086] Record the actual daily wear and tear of the carbon skateboard. ,like >2 If the wear rate of the carbon skateboard exceeds 7 consecutive days, it is marked as abnormal, characterized by excessively fast wear.

[0087] like >3 If the wear rate of the carbon skateboard exceeds 3 consecutive days, it is marked as abnormal, characterized by excessively rapid wear.

[0088] like < If the wear rate of the carbon skateboard exceeds 7 consecutive days, it is marked as abnormal, characterized by excessively slow wear.

[0089] like < If the wear rate of the carbon skateboard exceeds 3 consecutive days, it is marked as abnormal, characterized by severely slow wear.

[0090] This invention significantly improves the efficiency of carbon skid plate thickness detection. Taking a three-car maglev train as an example, if the train passes through the detection system at 5 km / h, the detection time is approximately 43 seconds; if the train passes through the detection system at 5 km / h, the detection time is approximately 21 seconds. The entire detection process is automated, requiring no on-site operation by maintenance personnel. In terms of maintenance effectiveness, the detection accuracy is high, and the detection results are stable, eliminating false detections and missed detections caused by manual inspection. It also provides carbon skid plate thickness warning and abnormal wear analysis functions, enabling a more comprehensive analysis of the carbon skid plate's health status and improving vehicle safety.

[0091] In one embodiment, the step of obtaining multiple sets of original point clouds corresponding to multiple structures to be detected in a maglev train current receiver includes: obtaining the current receiver point cloud corresponding to the maglev train current receiver; the maglev train current receiver includes multiple target structures to be detected; and segmenting the current receiver point cloud using a pre-set dividing line to obtain multiple sets of original point clouds each containing multiple structures to be detected.

[0092] In one embodiment, the step of segmenting the current receiver point cloud using pre-set dividing lines to obtain multiple original point cloud sets, each containing multiple structures to be detected, includes: constructing a spatial rectangular coordinate system; the X-axis of the spatial rectangular coordinate system is perpendicular to the direction of operation of the maglev train, and the Y-axis of the spatial rectangular coordinate system is parallel to the direction of operation of the maglev train; obtaining the edge point cloud of the current receiver point cloud in the X-axis direction, and obtaining two dividing lines based on the edge point cloud coordinates in the two edge point cloud sets, the physical parameters of each structure of the current receiver, and the spatial topology; segmenting the current receiver point cloud using the two dividing lines to obtain a first original point cloud set, a second original point cloud set, and a third original point cloud set, each containing multiple structures to be detected.

[0093] In one specific embodiment, such as Figure 3 The diagram illustrates a flowchart of a method for detecting a current collector in a maglev train. Before segmenting the corresponding original point cloud set using the point cloud model after each spatial transformation to obtain the corresponding optimized point cloud set, the method further includes: calculating the overlap degree between each original point cloud set and the corresponding registered point cloud set; if the overlap degree is not within a pre-set overlap degree threshold range, the detection result for the corresponding structure to be detected is output, indicating an anomaly; if the overlap degree is within a pre-set overlap degree threshold range, the mean of the overlapping region between each original point cloud set and the corresponding registered point cloud set is calculated; if the mean of the overlapping region is not within a pre-set mean threshold range, the detection result for the corresponding structure to be detected is output, indicating an anomaly; if the mean of the overlapping region is within a pre-set mean threshold range, the variance of the overlapping region between each original point cloud set and the corresponding registered point cloud set is calculated; if the variance of the overlapping region is not within a pre-set variance threshold range, the detection result for the corresponding structure to be detected is output, indicating an anomaly.

[0094] In this embodiment, a single original point cloud set is used as an example to calculate the spatially transformed point cloud model. With the original point cloud collection overlap The system determines whether the overlap ratio is within the overlap ratio threshold range. If it is not, it indicates that the overlap ratio between the collected point cloud and the template point cloud is low, and the program issues a warning that an anomaly has occurred in this area. If it is within the overlap ratio threshold range, the average value of the overlapping areas is calculated sequentially. With variance The program determines whether the mean and variance meet the requirements. If they do not, it means that although the collected point cloud and the template point cloud largely overlap, there may be deformed areas. The program issues an abnormal warning. Similarly, it obtains the overlap degree, the mean of the overlapping area, and the variance of the overlapping area of ​​other original point cloud sets, and determines whether they meet the requirements based on the corresponding thresholds. If they do not meet the requirements, an abnormal warning is issued.

[0095] In one embodiment, the step of segmenting the corresponding original point cloud set using each spatially transformed point cloud model to obtain the corresponding optimized point cloud set includes: segmenting the corresponding original point cloud set according to the boundary information of each spatially transformed point cloud model, extracting overlapping region point clouds, clustering the overlapping region point clouds, removing point cloud clusters with a small number of neighboring point clouds, and obtaining the corresponding optimized point cloud set.

[0096] In one embodiment, the step of acquiring multiple target structure images corresponding to multiple structures to be detected includes: acquiring the image to be detected of the maglev train current collector; and processing the image to be detected using a contour matching method to obtain multiple target structure images corresponding to multiple structures to be detected.

[0097] In one embodiment, the method further includes: acquiring the train speed and the fluctuation amount perpendicular to the train's direction of travel at each moment; determining whether the image to be detected has motion blur based on the relationship between the fluctuation amount corresponding to the fluctuation time and the image accuracy; when the image has motion blur, adjusting the current camera's exposure time according to the train's current speed, and updating the image to be detected based on the adjusted exposure time.

[0098] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0099] In one embodiment, such as Figure 9 As shown, a magnetic levitation train current collector detection device is provided, including: a data acquisition module 902, a point cloud optimization module 904, a straight line fitting module 906, and a structure detection module 908, wherein:

[0100] Data acquisition module 902 is used to acquire multiple sets of original point clouds and multiple target structure images corresponding to multiple structures to be detected in the current collector of the maglev train;

[0101] The point cloud optimization module 904 is used to perform point cloud registration between each original point cloud set and the point cloud model of the corresponding structure to be detected, so as to obtain the corresponding spatially transformed point cloud model. The corresponding original point cloud set is segmented using each spatially transformed point cloud model to obtain the corresponding optimized point cloud set.

[0102] The straight line fitting module 906 is used to find two edge point cloud sets along the vertical direction of the maglev train in each optimized point cloud set, obtain the first straight line corresponding to each optimized point cloud set based on the edge point cloud coordinates in the two edge point cloud sets, find the edge pixel position along the vertical direction of the maglev train in each target structure image, and obtain the second straight line corresponding to the target structure image based on the edge pixel position.

[0103] The structure detection module 908 is used to obtain the slope threshold range based on the slope of the second straight line corresponding to the second straight line and the corresponding threshold, and to detect whether multiple structures to be detected of the current receiver are abnormal based on the relationship between the slope of the first straight line corresponding to each optimized point cloud set and the slope threshold range.

[0104] In one embodiment, the data acquisition module 902 is further used to acquire the current receiver point cloud corresponding to the current receiver of the maglev train; the current receiver of the maglev train includes multiple target structures to be detected; the current receiver point cloud is segmented using a pre-set dividing line to obtain multiple original point cloud sets, each containing multiple target structures to be detected.

[0105] In one embodiment, the point cloud optimization module 904 is further used to construct a spatial rectangular coordinate system; the X-axis of the spatial rectangular coordinate system is perpendicular to the direction of operation of the maglev train, and the Y-axis of the spatial rectangular coordinate system is parallel to the direction of operation of the maglev train; the edge point cloud of the current receiver point cloud in the X-axis direction is obtained, and two dividing lines are obtained based on the edge point cloud coordinates in the two edge point cloud sets, the physical parameters of each structure of the current receiver, and the spatial topology relationship; the current receiver point cloud is divided using the two dividing lines to obtain a first original point cloud set, a second original point cloud set, and a third original point cloud set, each containing multiple structures to be detected.

[0106] In one embodiment, the method further calculates the overlap degree between each original point cloud set and the corresponding registered point cloud set. If the overlap degree is not within a preset overlap degree threshold range, the detection result for the corresponding structure to be detected is output; the detection result is abnormal. If the overlap degree is within a preset overlap degree threshold range, the mean value of the overlapping region between each original point cloud set and the corresponding registered point cloud set is calculated. If the mean value of the overlapping region is not within a preset mean value threshold range, the detection result for the corresponding structure to be detected is output; the detection result is abnormal. If the mean value of the overlapping region is within a preset mean value threshold range, the variance of the overlapping region between each original point cloud set and the corresponding registered point cloud set is calculated. If the variance of the overlapping region is not within a preset variance threshold range, the detection result for the corresponding structure to be detected is output; the detection result is abnormal.

[0107] In one embodiment, the point cloud optimization module 904 is further configured to segment the corresponding original point cloud set according to the boundary information of the point cloud model after each spatial transformation, extract the overlapping area point cloud, cluster the overlapping area point cloud, remove the small number of point cloud clusters in the neighboring point cloud, and obtain the corresponding optimized point cloud set.

[0108] In one embodiment, the data acquisition module 902 is further used to acquire the image to be detected of the maglev train current collector; and to process the image to be detected using a contour matching method to obtain multiple target structure images corresponding to multiple structures to be detected.

[0109] In one embodiment, the data acquisition module 902 is further configured to acquire the train speed and the fluctuation amount perpendicular to the train running direction at each moment; determine whether the image to be detected has motion blur based on the relationship between the fluctuation amount corresponding to the fluctuation time and the image accuracy; when the image has motion blur, adjust the exposure time of the current camera according to the train running speed at the current moment, and update the image to be detected according to the adjusted exposure time.

[0110] Specific limitations regarding the maglev train current collector detection device can be found in the limitations of the maglev train current collector detection method described above, and will not be repeated here. Each module in the aforementioned maglev train current collector detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0111] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for detecting current collectors on a maglev train. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0112] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0113] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described above.

[0114] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0115] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0116] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for detecting the current collector of a maglev train, characterized in that, The method includes: Acquire multiple original point cloud sets and multiple target structure images corresponding to multiple structures to be detected in the current collector of a maglev train; Each original point cloud set is registered with the point cloud model of the corresponding structure to be detected to obtain the corresponding spatially transformed point cloud model. The corresponding original point cloud set is then segmented using each spatially transformed point cloud model to obtain the corresponding optimized point cloud set. Find two edge point cloud sets along the vertical direction of the maglev train in each optimized point cloud set. Based on the edge point cloud coordinates in the two edge point cloud sets, obtain the first straight line corresponding to each optimized point cloud set. Find the edge pixel position along the vertical direction of the maglev train in each target structure image. Based on the edge pixel position, obtain the second straight line corresponding to the target structure image. Based on the slope of the second line corresponding to the second line and the corresponding threshold, the slope threshold range is obtained. Based on the relationship between the slope of the first line corresponding to each optimized point cloud set and the slope threshold range, the abnormality of the multiple structures to be detected in the current receiver is detected.

2. The method according to claim 1, characterized in that, The steps for obtaining multiple sets of original point clouds corresponding to multiple structures to be detected in the current collector of a maglev train include: Obtain the point cloud of the current collector corresponding to the current collector of the maglev train; the current collector of the maglev train includes multiple target structures to be detected; The current receiver point cloud is segmented using pre-set dividing lines to obtain multiple original point cloud sets, each containing the multiple structures to be detected.

3. The method according to claim 2, characterized in that, The step of segmenting the current receiver point cloud using pre-set dividing lines to obtain multiple original point cloud sets, each containing the multiple structures to be detected, includes: Construct a spatial rectangular coordinate system; the X-axis of the spatial rectangular coordinate system is perpendicular to the direction of travel of the maglev train, and the Y-axis of the spatial rectangular coordinate system is parallel to the direction of travel of the maglev train; Obtain the edge point cloud of the current receiver point cloud in the X-axis direction, and obtain two dividing lines based on the edge point cloud coordinates in the two edge point cloud sets, the physical parameters of each structure of the current receiver, and the spatial topology. The current receiver point cloud is segmented using the two dividing lines to obtain a first original point cloud set, a second original point cloud set, and a third original point cloud set, each containing the multiple structures to be detected.

4. The method according to claim 1, characterized in that, Before segmenting the corresponding original point cloud set using the point cloud model after each spatial transformation to obtain the corresponding optimized point cloud set, the process also includes: Calculate the overlap degree between each original point cloud set and the corresponding registered point cloud set. If the overlap degree is not within the preset overlap degree threshold range, output the detection result of the corresponding structure to be detected; the detection result is an anomaly. If the overlap is within a preset overlap threshold range, the average overlap region between each original point cloud set and the corresponding registered point cloud set is calculated. If the average overlap region is not within a preset average threshold range, the detection result of the corresponding structure to be detected is output; the detection result is an anomaly. If the mean of the overlapping region is within a preset mean threshold range, calculate the variance of the overlapping region between each original point cloud set and the corresponding registered point cloud set. If the variance of the overlapping region is not within a preset variance threshold range, output the detection result of the corresponding structure to be detected; the detection result is abnormal.

5. The method according to claim 1, characterized in that, The steps of segmenting the corresponding original point cloud set using the point cloud model after each spatial transformation to obtain the corresponding optimized point cloud set include: Based on the boundary information of each spatially transformed point cloud model, the corresponding original point cloud set is segmented, overlapping region point clouds are extracted, and the overlapping region point clouds are clustered to remove the small number of point cloud clusters in the nearest neighbor point clouds, thereby obtaining the corresponding optimized point cloud set.

6. The method according to claim 1, characterized in that, The steps of obtaining multiple target structure images corresponding to the multiple structures to be detected include: Acquire the image to be tested from the current collector of the maglev train; The image to be detected is processed using a contour matching method to obtain multiple target structure images corresponding to the multiple structures to be detected.

7. The method according to claim 6, characterized in that, The method further includes: The train speed and fluctuation in the direction perpendicular to the train's direction of travel are obtained at each moment during the operation of the maglev train. Based on the relationship between the fluctuation amount corresponding to the fluctuation time and the image precision, it is determined whether the image to be detected has ghosting; When image motion blur is present, the exposure time of the current camera is adjusted according to the current train speed, and the image to be detected is updated according to the adjusted exposure time.

8. A detection device for a current collector of a maglev train, characterized in that, The device includes: The data acquisition module is used to acquire multiple sets of original point clouds and multiple target structure images corresponding to multiple structures to be detected in the current collector of the maglev train; The point cloud optimization module is used to perform point cloud registration between each original point cloud set and the point cloud model of the corresponding structure to be detected, so as to obtain the corresponding spatially transformed point cloud model. The corresponding original point cloud set is segmented using each spatially transformed point cloud model to obtain the corresponding optimized point cloud set. The straight line fitting module is used to find two edge point cloud sets along the vertical direction of the maglev train in each optimized point cloud set, obtain the first straight line corresponding to each optimized point cloud set based on the edge point cloud coordinates in the two edge point cloud sets, find the edge pixel position along the vertical direction of the maglev train in each target structure image, and obtain the second straight line corresponding to the target structure image based on the edge pixel position. The structure detection module is used to obtain a slope threshold range based on the slope of the second straight line corresponding to the second straight line and the corresponding threshold, and to detect whether the multiple structures to be detected of the current receiver are abnormal based on the relationship between the slope of the first straight line corresponding to each optimized point cloud set and the slope threshold range.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images