A method and system for online monitoring of a power shoe

Abstract

The application belongs to the technical field of rail transit detection, and particularly relates to a power shoe online monitoring method and system. The power shoe online monitoring method comprises the following steps: real-time detection of train running information, generation of a primary trigger signal according to the train running information, control of a three-dimensional line scanning camera to collect three-dimensional point cloud data of the power shoe according to the primary trigger signal, wherein the train running information comprises a train coming direction and a train running speed; generation of a secondary trigger signal according to the three-dimensional point cloud data, control of a two-dimensional color area array camera to capture the power shoe according to the secondary trigger signal, and collection of a two-dimensional color image of the power shoe; fusion processing and identification of the three-dimensional point cloud data and the two-dimensional color image, and calculation of geometric parameters and judgment of a surface state of the power shoe according to a result of the fusion processing and identification.

Description

Technical Field

[0001] This application belongs to the field of rail transit detection technology, and in particular relates to an online monitoring method and system for electric receiving shoes. Background Technology

[0002] The current collector shoe (also known as a current collector or current collector) is a key current collection component for urban rail transit trains to obtain power from the third rail. The wear, geometric deviation, and surface damage (such as burning or chipping) of its carbon sliding plate directly affect the safe and stable current collection of the train. To ensure operational safety, the condition of the current collector shoe must be inspected regularly.

[0003] Currently, the identification of geometric parameters and fault conditions of power receiving shoes mainly involves scanning the carbon sliding plate with a 2D line scan camera or capturing images of the carbon sliding plate with a 2D area scan camera, and then applying 2D image recognition algorithms to identify the power receiving shoe. This type of method is typically based on grayscale and brightness recognition, which requires precise lighting and performs poorly on images with blurred grayscale distributions. Furthermore, for the identification of carbon sliding plate surface burns, since surface burns are generally dark orange in color and vary in size and shape, they are similar in color to the carbon sliding plate surface in grayscale images, making it difficult to identify boundaries or perform morphological analysis. In addition, existing methods for measuring the geometric dimensions of carbon sliding plates also convert the number of pixels occupied by the sides of the carbon sliding plate in a 2D grayscale image into its size. This method is often limited by actual operating conditions and cannot achieve the required accuracy.

[0004] Therefore, there is an urgent need for a method to monitor electric boots to improve the accuracy of calculating the geometric parameters of electric boots and identifying their surface conditions. Summary of the Invention

[0005] This invention addresses at least some of the aforementioned technical problems, providing a method and system for online monitoring of electric receiving boots. By integrating three-dimensional geometric morphology with two-dimensional color texture information, it achieves high-precision measurement of the geometric dimensions of the electric receiving boot and reliable identification of various surface anomalies, effectively overcoming the problems of measurement errors caused by posture changes, sensitivity to illumination, and lack of depth information in two-dimensional detection.

[0006] The first aspect of the embodiments of this disclosure provides an online monitoring method for power receiving shoes, comprising: The system monitors train travel information in real time, generates an initial trigger signal based on the train travel information, and controls a 3D line scan camera to collect 3D point cloud data of the power receiving shoe based on the initial trigger signal; the train travel information includes the train's direction of travel and train speed. A secondary trigger signal is generated based on the three-dimensional point cloud data, and a two-dimensional color area array camera is controlled based on the secondary trigger signal to capture the electric receiving shoe and acquire a two-dimensional color image of the electric receiving shoe. The three-dimensional point cloud data and the two-dimensional color image are fused and recognized. Based on the results of the fusion and recognition, the geometric parameters of the power receiving shoe are calculated and the surface condition is judged.

[0007] The technical solution provided in this application brings at least the following beneficial effects: based on the precise positioning of three-dimensional point cloud to trigger two-dimensional image capture, by integrating three-dimensional geometric data and two-dimensional color information for analysis and processing, posture error is eliminated and measurement accuracy is improved, while geometric parameter measurement and surface damage identification are realized, thus achieving non-contact dynamic online monitoring of the power receiving shoe.

[0008] In other embodiments of this application, the fusion processing and recognition of the three-dimensional point cloud data and the two-dimensional color image includes an image preprocessing step, which includes: The three-dimensional point cloud data is subjected to iterative nearest point (ICP) registration to obtain the first point cloud data aligned with the standard template point cloud; Extract the contour data of the power receiving boot from the first point cloud data; The contour data is fitted with a B-spline curve to obtain a contour curve function; Calculate the difference value of the contour curve function, and locate the contour inflection point of the electric receiving shoe based on the difference value; Two spatial straight lines are obtained by performing linear fitting on the data on both sides of the contour inflection point; The geometric parameters of the power receiving boot are calculated and the surface condition is determined based on the spatial straight line fitted by the straight line.

[0009] The technical solution provided in this application has at least the following beneficial effects: by registering the three-dimensional point cloud to eliminate attitude deviation, then extracting the contour and fitting the curve, and using the difference method to accurately locate the inflection point and perform straight line fitting, a high-precision data benchmark is provided for geometric parameter calculation and state judgment, which significantly improves the reliability and accuracy of measurement.

[0010] In other embodiments of this application, the fusion processing and recognition of the three-dimensional point cloud data and the two-dimensional color image, and the calculation of the geometric parameters and surface condition judgment of the power receiving shoe based on the results of the fusion processing and recognition, include a surface burn judgment step, which includes: Locate the electric boot region in the two-dimensional color image and convert the electric boot region into the Lab color space; The texture features of the power-receiving shoe region are extracted in the Lab color space, and anomaly detection is performed. When an abnormal region is detected by the anomaly detection, image segmentation and morphological processing are performed on the abnormal region to obtain at least one connected component. Calculate the area and shape factor of each connected region. When the area of ​​the connected region is greater than or equal to a first area threshold and the shape factor is within a preset range, it is determined that the electric boot has surface burn damage.

[0011] The technical solution provided in this application brings at least the following beneficial effects: by converting the Lab color space and extracting texture features, it effectively eliminates light interference, and by combining morphological processing and geometric feature determination, it accurately identifies surface burn damage with varying shapes.

[0012] In other embodiments of this application, the fusion processing and recognition of the three-dimensional point cloud data and the two-dimensional color image, and the calculation of the geometric parameters and surface condition judgment of the power receiving shoe based on the fusion processing and recognition results, include a surface spalling judgment step, which includes: The three-dimensional point cloud data is registered and aligned with the standard template point cloud to obtain the first three-dimensional point cloud data; Calculate the volume difference between the first 3D point cloud data and the standard template point cloud; When the volume difference is greater than a preset threshold, it is determined that there is surface chipping damage.

[0013] The technical solution provided in this application brings at least the following benefits: Based on the comparative analysis of the volume difference of three-dimensional point clouds, quantitative detection of surface chipping damage is realized, which effectively makes up for the limitation that two-dimensional images cannot obtain depth information and improves the accuracy of damage assessment.

[0014] In other embodiments of this application, calculating the volume difference between the first three-dimensional point cloud data and the standard template point cloud includes: Construct a three-dimensional distance field and calculate the difference of the signed distance function (SDF) between the target point cloud and the standard template point cloud; The volume difference is calculated by discretized integration based on the difference in the symbolic distance function (SDF).

[0015] The technical solution provided in this application brings at least the following beneficial effects: by constructing the difference between the three-dimensional distance field and the symbolic distance function and performing discrete integration, quantitative analysis of the depth of the dropped block is achieved, which significantly improves the accuracy and reliability of damage detection.

[0016] In other embodiments of this application, the contour inflection points of the power receiving boot include an upper edge inflection point and a lower edge inflection point. The step of calculating the geometric parameters and determining the surface condition of the power receiving boot based on the fusion processing and recognition results includes a thickness calculation step, which includes: Based on the inflection point of the upper edge contour of the power receiving shoe, two spatial straight lines are fitted on both sides of the inflection point of the upper edge contour. Calculate the intersection point of the two spatial lines; Based on the upper edge contour inflection point, draw a vertical straight line along the height direction of the power receiving shoe to the lower edge to obtain the first lower edge inflection point. Calculate the spatial distance between the intersection point and the first lower edge inflection point, and use the spatial distance as the thickness value of the power receiving shoe. The upper edge inflection point of the power receiving shoe refers to the transition point on the upper surface of the power receiving shoe at the junction of the carbon sliding plate and the bracket on the contour curve; the lower edge inflection point of the power receiving shoe refers to the lowest point of the bottom of the power receiving shoe on the contour curve.

[0017] The technical solution provided in this application brings at least the following beneficial effects: by fitting a straight line to find the intersection point, the thickness measurement reference point is accurately located, and the thickness value is accurately calculated based on the three-dimensional spatial distance, effectively eliminating the projection error caused by the posture deviation.

[0018] In other embodiments of this application, the fusion processing and recognition of the three-dimensional point cloud data and the two-dimensional color image, and the calculation of the geometric parameters and surface condition judgment of the power receiving boot based on the results of the fusion processing and recognition, include a height calculation step, which includes: The upper and lower edge inflection points of the power receiving shoe are obtained through image preprocessing; Based on the known fixed installation angle of the camera, the calibrated distance from the camera center to the rail surface, and the coordinates of the upper and lower edge inflection points, spatial trigonometric geometric relationships are calculated to obtain the vertical height value of the power receiving shoe.

[0019] The technical solution provided in this application brings at least the following benefits: based on spatial trigonometric geometry, the camera coordinates are converted into vertical height with the track surface as a reference, eliminating attitude errors and realizing non-contact precision measurement.

[0020] In other embodiments of this application, the fusion processing and recognition of the three-dimensional point cloud data and the two-dimensional color image, and the calculation of the geometric parameters and surface condition judgment of the power receiving boot based on the results of the fusion processing and recognition, include a length calculation step, which includes: The coordinates of the left endpoint of the upper edge and the right endpoint of the upper edge of the electric boot are obtained through image preprocessing. Calculate the spatial Euclidean distance between the left endpoint of the upper edge and the right endpoint of the upper edge, and use the spatial Euclidean distance as the length value of the power receiving shoe.

[0021] The technical solution provided in this application has at least the following beneficial effects: by directly calculating the spatial Euclidean distance using three-dimensional coordinates, the actual measurement of the length of the power receiving shoe is realized, effectively avoiding the errors caused by two-dimensional projection deformation.

[0022] In other embodiments of this application, the fusion processing and recognition of the three-dimensional point cloud data and the two-dimensional color image, and the calculation of the geometric parameters and surface condition judgment of the power receiving boot based on the results of the fusion processing and recognition, include a tilt angle calculation step, which includes: Obtain the horizontal reference direction in the camera coordinate system; The coordinates of the upper edge endpoint of the power receiving shoe are obtained through image preprocessing, and the direction of the upper edge line of the power receiving shoe in the camera coordinate system is calculated. Calculate the angle between the direction of the upper edge straight line and the horizontal reference direction, and use it as the tilt angle of the power receiving shoe.

[0023] The technical solution provided in this application has at least the following beneficial effects: based on the horizontal reference in the camera coordinate system and the measured upper edge direction, the tilt angle is accurately quantified by calculating the included angle, providing a reliable basis for the attitude evaluation of the power receiving shoe.

[0024] A second aspect of the present disclosure provides an online monitoring system for a power receiving shoe, used to implement the online monitoring method for a power receiving shoe disclosed in the first aspect of the present disclosure, comprising: A 3D line scan camera is configured to acquire 3D point cloud data of the power receiving shoe; A two-dimensional color area array camera is configured to acquire two-dimensional color images of the power receiving shoe; The upper-level fusion and recognition module, connected to the 3D line scan camera and the 2D color area scan camera, is configured to receive the 3D point cloud data and the 2D color image, and perform fusion processing and recognition. The control module, connected to the 3D line scan camera, the 2D color area scan camera, and the upper-level fusion and recognition module, is configured as follows: The system monitors train travel information in real time, generates an initial trigger signal based on the train travel information, and controls the 3D line scan camera to collect 3D point cloud data of the power receiving shoe based on the initial trigger signal; the train travel information includes the train travel direction and the train travel speed. A secondary trigger signal is generated based on the three-dimensional point cloud data, and the two-dimensional color area array camera is controlled to capture the electric receiving shoe based on the secondary trigger signal to acquire a two-dimensional color image of the electric receiving shoe; The upper-level fusion and recognition module is controlled to perform fusion processing and recognition on the three-dimensional point cloud data and the two-dimensional color image, and to calculate the geometric parameters and determine the surface condition of the power receiving shoe based on the results of the fusion processing and recognition.

[0025] Compared with the prior art, the present invention has the following beneficial effects: (1) By fusing three-dimensional point cloud data with two-dimensional color images, the problems of projection error and lack of depth information caused by posture change in traditional two-dimensional detection are overcome, and high-precision measurement of the geometric parameters of the power receiving shoe and reliable identification of surface damage are realized. (2) By generating secondary trigger signals based on three-dimensional point cloud analysis to control two-dimensional capture, the spatiotemporal synchronous acquisition of three-dimensional and two-dimensional data was realized, providing a high-quality data foundation for fusion analysis.

[0026] The above description is merely an overview of the technical solution disclosed herein. In order to better understand the technical means of this disclosure and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this disclosure more apparent and understandable, specific embodiments of this disclosure are described below. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a schematic flowchart of the online monitoring method for the power receiving shoe according to an embodiment of this application; Figure 2 This is a topology diagram of the online monitoring system for the power receiving shoe according to an embodiment of this application; Figure 3 This is a schematic diagram of the installation of the power receiving shoe online monitoring system according to an embodiment of this application; Figure 4 This is a schematic diagram of point cloud data from an embodiment of the power receiving shoe online monitoring system of this application. Figure 1 ; Figure 5 This is a schematic diagram of point cloud data from an embodiment of the power receiving shoe online monitoring system of this application. Figure 2 ; Figure 6 This is a coordinate schematic diagram of the online monitoring system for the power receiving shoe according to an embodiment of this application; In the above figures, 1. First upper surface detection module; 2. Second upper surface detection module; 3. First lower surface detection module; 4. Second lower surface detection module. Detailed Implementation

[0029] To make the technical problems, technical solutions, and beneficial effects to be solved by 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 are not intended to limit the scope of this application.

[0030] The prefixes such as "first" and "second" used in this application embodiment are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this application embodiment does not constitute a limitation on the described objects. The description of the described objects is given in the claims or the context of the embodiments, and should not constitute unnecessary restrictions due to the use of such prefixes. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0031] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. In the description of the embodiments of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B; the term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone.

[0032] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0033] In this application, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0034] Currently, the identification of geometric parameters and fault status of power receiving shoes mainly involves scanning the carbon receiving plate with a 2D line scan camera or capturing images of the carbon receiving plate with a 2D area scan camera to obtain its image, and then applying 2D image recognition algorithms to identify the power receiving shoe. For example, the method described in CN 116067282 A uses a black-and-white area scan camera to capture images of the power receiving device and then applies traditional algorithms, namely: first, deep learning is used to locate the position of the power receiving device in the image, and then methods such as smoothing, edge extraction, and grayscale enhancement are used for further identification. This type of method is usually based on grayscale and brightness recognition, which has high requirements for lighting and poor recognition performance on images with blurred grayscale distribution. At the same time, for the identification of carbon receiving plate surface burns, since surface burns are generally dark orange in color and vary in shape and size, they are similar in color to the carbon receiving plate surface in grayscale images, making it difficult to identify boundaries or perform morphological analysis. CN 115439406 A describes the application of grayscale image deep learning to identify loose nuts, detached nuts, cracks in carbon slide plates, and defects in carbon slide plates. This method still relies on two-dimensional grayscale images for identification. For faults like cracks or defects in carbon slide plates, identification is limited to morphological analysis, lacking depth data (i.e., the depth of the crack or the depth of the missing piece). Furthermore, existing methods for measuring the geometric dimensions of carbon slide plates also convert the number of pixels occupied by the sides of the carbon slide plate in a two-dimensional grayscale image into its size. This method has the following drawbacks: 1. This method usually requires the relative position of the side of the carbon slide plate being measured to remain unchanged with respect to the camera target surface. However, in actual working conditions, the carbon slide plate is subjected to the action of torsion springs and swing arms, and its tilt angle often deviates. In this case, the 2D camera collects the projection of the side of the carbon slide plate in the camera rather than the actual side height, which will result in a large error in the measured geometric parameters.

[0035] 2. Edge extraction from 2D grayscale images often requires a certain gradient. However, in actual working conditions, the carbon slide plate and the background are basically black, and there is often a deviation of a few pixels at the edges. At the same time, due to the limitations of vehicle equipment clearance, the camera is often installed at a large distance from the power receiving shoe, which means a large depth of field. Even a single-digit pixel deviation can bring millimeter-level errors, which cannot meet the process requirements.

[0036] To address the above problems, this application provides an online monitoring method and system for power receiving shoes. The specific embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0037] The first aspect of the embodiments of this disclosure provides an online monitoring method for power receiving shoes, such as... Figure 1 As shown, the method includes: S01: Real-time detection of train travel information, generation of initial trigger signal based on the train travel information, and control of the three-dimensional line scan camera to collect three-dimensional point cloud data of the power receiving shoe based on the initial trigger signal; train travel information includes train direction of travel and train speed.

[0038] For example, when a train enters the detection area, the through-beam laser sensor is interrupted by the wheels or the vehicle body, generating a first trigger signal. The control module confirms the train's direction based on the first trigger signal and activates the speed-measuring radar to collect the train's speed in real time. The control module calculates the delay time based on a preset acquisition start position (usually the moment the power receiving shoe is about to enter the camera's field of view) and the current vehicle speed to generate an initial trigger signal. The initial trigger signal is output to the 3D line scan camera in RS422 differential level form, triggering it to begin continuous line scanning acquisition. The 3D line scan camera uses the laser triangulation principle, outputting a contour line (containing 4096 points) with each scan. As the train moves, it accumulates 3D point cloud data covering the entire power receiving shoe.

[0039] S02: Generate a secondary trigger signal based on the 3D point cloud data, and control the 2D color area array camera to capture the power receiving shoe and acquire a 2D color image of the power receiving shoe based on the secondary trigger signal.

[0040] For example, the point cloud data stream output in real time by the 3D line scan camera is sent to the buffer of the upper-level fusion and recognition module. The control module performs a simple height threshold judgment based on the point cloud data: when a sufficient number of point clouds appear in multiple consecutive contours within a preset depth of field, it is determined that the receiving shoe has entered the center area of ​​the camera's field of view. At this time, the control module precisely delays the timing based on the current vehicle speed to ensure that the receiving shoe moves to the ideal shooting position of the 2D camera and outputs a secondary trigger signal to the 2D color area scan camera. After receiving the secondary trigger signal, the 2D color area scan camera captures a high-definition color image (for example, 5 megapixels) using global exposure and uploads the image to the upper-level fusion and recognition module.

[0041] S03: Perform fusion processing and recognition on 3D point cloud data and 2D color images, and calculate the geometric parameters and surface condition of the power receiving shoe based on the results of fusion processing and recognition.

[0042] For example, iterative nearest point (ICP) registration is performed on 3D point cloud data to eliminate attitude deviations, contours are extracted and curves are fitted, inflection points are accurately located using the finite difference method and then straight lines are fitted. Based on the fitting results, geometric parameters such as the thickness, height, length and tilt angle of the electric boot are accurately calculated. At the same time, the 2D color image is converted to the Lab color space to extract texture features for anomaly detection. Surface burning damage is identified through morphological processing and connected component analysis, and surface chipping defects are detected by comparing the volume difference between the registered point cloud and the standard template. This achieves deep fusion and comprehensive recognition of 3D geometric shape and 2D color texture information.

[0043] This application's technical solution generates an initial trigger signal by real-time detection of train travel information, controlling a 3D line scan camera to dynamically acquire point cloud data from the power receiving shoe, ensuring synchronization between data acquisition and train operation. Then, based on 3D point cloud analysis, a secondary trigger signal is generated to precisely control a 2D color area array camera for image capture, achieving spatiotemporal collaborative acquisition of 3D and 2D data. In the fusion processing stage, the 3D point cloud is used to eliminate attitude deviations and accurately calculate geometric parameters. This method achieves non-contact dynamic online monitoring, effectively overcoming the limitations of traditional 2D detection which is susceptible to attitude changes and lighting conditions, thus improving measurement accuracy and the reliability of fault identification.

[0044] In a specific illustrative embodiment, the fusion processing and recognition of three-dimensional point cloud data and two-dimensional color images includes image preprocessing, which includes the following steps: S301: Perform iterative nearest point (ICP) registration on the 3D point cloud data to obtain the first point cloud data aligned with the standard template point cloud; Specifically, the point cloud data acquired by the 3D line scan camera is filtered and then matched with a standard template point cloud using the Iterative Closest Point (ICP) point cloud registration method to correct the position of the point cloud in space. The initial input includes two frames of original point cloud data, an initial estimate of the transformation, and a criterion for stopping the iteration. Then, the rigid body transformations (translation and rotation) of the two original point clouds are iteratively corrected to minimize the distance between all point sets, and finally the first point cloud data after transformation is output.

[0045] S302: Extract the contour data of the power receiving shoe from the first point cloud data; For example, from the registered point cloud data, a contour of the power receiving shoe is extracted at fixed intervals along the train's direction of travel. The fixed intervals can be set according to the specific point cloud data, and data from the contour is obtained. The median filtering algorithm was then used to remove outliers and extract the useful curve segments to obtain the processed contour data. (Based on analysis of the design drawings, the inflection point is located approximately 1 / 10 of the collected point cloud data; therefore, the first 500 points are directly selected.) Each data point... Represents the coordinates of the i-th sampling point on the extracted contour curve; This represents the displacement value along the lateral direction (e.g., the width direction) of the power receiving shoe. This is the height value (or distance relative to the reference plane) corresponding to that point.

[0046] S303: Fit B-spline curves to the contour data to obtain the contour curve function; refer to Figure 4For example, a cubic B-spline curve is fitted to the truncated discrete points to obtain a smooth contour curve function f(x).

[0047] S304: Calculate the difference value of the profile curve function, and locate the profile inflection point of the electric receiving shoe based on the difference value; For example, for a curvilinear function f(x), calculate the difference: ;in, The x-coordinate of the first point taken when discretizing the fitted contour curve function f(x); h represents the x-coordinate increment (i.e., interval) between two adjacent sampling points. This represents the x-coordinate of the k-th sampling point.

[0048] Judge the difference value in sequence Find the first difference value that approaches 0.001, which is the first inflection point of the profile curve extracted by the power receiving shoe.

[0049] like Figure 4 As shown in the figure, the red mark indicates an inflection point on the profile curve of the power receiving shoe.

[0050] Repeat the process for all contours to obtain a series of inflection points, and take their average position in space as the final inflection point coordinates.

[0051] S305: Perform linear fitting on the data on both sides of the contour inflection point to obtain two spatial straight lines; For example, the point cloud data on both sides of the inflection point is divided into two point sets: the left point set (the working surface of the carbon skateboard) and the right point set (the bracket part). The RANSAC algorithm is used to fit three-dimensional spatial lines to the two point sets respectively, resulting in two spatial lines L1 and L2. Each line is represented by its direction vector and a point on the line. The RANSAC parameters are set as follows: 1) Objective function: This refers to the process of maximizing the number of interior points n under different iterations of the model parameters through m iterations. The objective function of the RANSAC algorithm applied to spatial line fitting is to obtain the parametric model containing the most interior points based on a known distance threshold K of the spatial line.

[0052] 2) Sampling subset size. During iterative calculations, model parameters need to be calculated using a subset, and each sampling subset should be the smallest sampling set with a data size of N. Spatial line fitting requires at least two spatial points; therefore, when the RANSAC algorithm is applied to three-dimensional spatial line fitting, N=2 is taken as the iteration termination condition. The number of iterations m for RANSAC can be theoretically calculated. Under a confidence level of P, during the loop, at least one sampling ensures that all N points in the sampling subset are interior points, thus guaranteeing that at least one sampling during the iteration process yields the maximum value of the objective function. Therefore, the loop termination conditions m and n should satisfy the following condition:

[0053] In the above conditional expressions: the confidence level P is generally set within the range of [95%, 99%]; in general, t, as the probability of an inlier in the dataset, is an unknown parameter. Therefore, the proportion of inliers under the worst-case condition can be taken, and then the current maximum proportion of inliers can be continuously updated as the number of iterations increases. Preferably, the confidence level P is set to 99%.

[0054] S306: Calculate the geometric parameters and determine the surface condition of the power receiving shoe based on the spatial straight line after straight line fitting.

[0055] For example, based on the fitted spatial straight line, geometric parameters such as the thickness, height, length, and tilt angle of the electric receiving shoe can be calculated, and the identification and judgment of chipping and burning on the surface of the electric receiving shoe can be performed.

[0056] This application's technical solution effectively eliminates spatial position deviations introduced by changes in the electric receiving shoe's attitude (such as tilting and swaying) during train operation by performing ICP registration on 3D point cloud data, ensuring the consistency between the measurement benchmark and the standard template. Furthermore, the contour is extracted from the registered point cloud and fitted with a B-spline curve. The inflection points of the contour are precisely located using the finite difference method, achieving accurate identification of key structural features of the electric receiving shoe. Fitting the data on both sides of the inflection point yields two spatial straight lines, providing a precise linear benchmark for calculating geometric parameters such as thickness, height, and angle. This method overcomes the projection errors caused by attitude changes in traditional 2D detection, significantly improving the accuracy and reliability of geometric parameter measurements.

[0057] In a specific illustrative embodiment, three-dimensional point cloud data and two-dimensional color images are fused and recognized. Based on the results of the fusion and recognition, the geometric parameters of the power receiving shoe are calculated and the surface condition is judged, including surface burn judgment. The surface burn judgment steps include: S311: Locate the electric boot region in a two-dimensional color image and convert the electric boot region to the Lab color space; Specifically, a multi-scale normalized cross-correlation template matching algorithm is used to locate the electric shoe region in the image, with a positioning error ≤ 3 pixels. The image of this region is then converted from the RGB color space to the Lab color space, and the luminance information is separated using the L channel while retaining the color information of the a and b channels.

[0058] S312: Extract texture features of the power-receiving shoe region in the Lab color space and perform anomaly detection; Specifically, Local Binary Pattern (LBP) texture features are extracted from channels a and b in the Lab space to form a feature vector. The isolated forest unsupervised anomaly detection algorithm is then used to score this feature vector.

[0059] S313: When an abnormal region is detected by anomaly detection, the abnormal region is segmented and morphologically processed to obtain at least one connected component; When the abnormal probability of the feature vector is greater than 0.7, it is determined that there is an abnormal region in the area, triggering the burn determination process.

[0060] For abnormal regions, the image is converted back to RGB space, and the red channel features are extracted (since burns are usually dark red), constructing a red feature map. Adaptive OTSU thresholding is applied to this map to obtain a binary image. Morphological processing is then performed: first, a 3×3 rectangular kernel is used to perform an erosion operation on the binary image to remove small noise points; then, a 5×5 rectangular kernel is used to perform two dilation operations to connect the broken burn regions. At least one connected component is found in the resulting image.

[0061] S314: Calculate the area and shape factor of each target connected region in the image, and determine the surface burn damage of the electric boot by double threshold screening. The specific judgment rule is: when the area of ​​the connected region is greater than or equal to the first area threshold and the shape factor is within a preset reasonable range, it is determined that the electric boot has surface burn damage.

[0062] This application's technical solution is the first to extract texture features and detect abnormal areas in the target area of ​​the power receiving boot within the Lab color space. Leveraging the independent nature of the luminance and color channels in the Lab color space, it completely separates illumination interference from target texture information, effectively overcoming the interference of uneven lighting and localized reflections in the on-site working environment on the accurate identification of burn marks. After preprocessing the initially screened abnormal areas with image segmentation, morphological closing operations, and denoising, the pixel area and shape factor of each connected region are further calculated to achieve a secondary, accurate screening of real damage and interference items. Burn patches on the surface of the power receiving boot often exhibit elliptical, regular patch characteristics in imaging. Combining the area threshold and shape factor as joint judgment criteria, it can effectively distinguish real burn damage from surface stains, localized reflections, dust adhesion, and other interference items, significantly improving the detection accuracy of surface burn defects and its robustness to on-site working conditions.

[0063] Specifically, calculate the area S (number of pixels) and shape factor of each connected component.

[0064] in, The total length of the closed outer contour of the target connected region, expressed in pixels, is calculated by traversing consecutive pixels along the edges of the connected region using a contour tracking algorithm. It is a core morphological parameter characterizing the extension and regularity of the connected region's contour. S represents the total number of pixels contained in the connected region, directly reflecting the size of the burned patch. When the connected region simultaneously satisfies… and When the system determines that the connected region is the actual area of ​​burn damage on the surface of the power receiving boot, it outputs a corresponding damage alarm signal.

[0065] In a specific illustrative embodiment, 3D point cloud data and 2D color images are fused and recognized. Based on the results of the fusion and recognition, the geometric parameters of the power receiving shoe are calculated and the surface condition is judged, including a surface spalling judgment step. The surface spalling judgment step includes: S321: Register and align the 3D point cloud data with the standard template point cloud to obtain the first 3D point cloud data; S322: Calculate the volume difference between the first 3D point cloud data and the standard template point cloud; For example, by constructing a three-dimensional distance field, the difference of the signed distance function (SDF) between the target point cloud and the template point cloud is calculated, and the volume difference is obtained by discrete integration.

[0066] This application's technical solution eliminates spatial position deviations caused by changes in the orientation of the power receiving shoe by registering and aligning real-time acquired 3D point clouds with standard template point clouds. The volume difference between the two is then calculated, extending the identification of debris damage from two-dimensional planar morphological analysis to three-dimensional spatial quantitative assessment. Compared to the limitations of traditional 2D images, which can only detect the projected area of ​​the defect, this method can obtain the depth information and true volume of the debris. When the volume difference exceeds a preset threshold, the damage can be accurately determined, effectively solving the problem that 2D detection cannot assess the severity of damage.

[0067] In a specific illustrative embodiment, calculating the volume difference between the first 3D point cloud data and the standard template point cloud includes: S3221: Construct a three-dimensional distance field and calculate the difference of the signed distance function (SDF) between the target point cloud and the standard template point cloud; For example, a uniform 3D mesh covering the power receiving shoe is constructed, and the shortest symbol distance to the standard template point cloud is calculated for each mesh, forming a distance field. The symbol distance of the target point cloud is calculated, and the difference between the two symbol distance fields is compared to obtain the symbol distance function difference. .

[0068] in, This represents the volume difference between the target point cloud and the standard template point cloud, i.e., the volume of any missing or damaged portion on the surface of the power receiving shoe (usually in mm³). Ω represents the integration region, typically referring to a three-dimensional spatial extent covering the entire power receiving shoe or a key area of ​​interest (e.g., a cuboid bounding box containing the entire carbon skateboard). The volume difference is calculated within this region. The signed distance function value of the target point cloud, representing the spatial point. The shortest signed distance to the surface of the power receiving boot point cloud after real-time acquisition and registration. It is generally stipulated that points inside an object are negative, those outside are positive, and those on the surface are zero. The signed distance function value is the standard template point cloud, representing the signed distance from the same spatial point to the surface of the standard template point cloud. A volume element represents a tiny unit of three-dimensional spatial integration.

[0069] S3222: Calculate the volume difference based on the discretized integral using the sign distance function (SDF) difference.

[0070] For example, the formula is calculated by discretizing the integral. Calculate the volume difference.

[0071] in, The sign distance difference flag for the i-th voxel (when...) Set to 1 if the condition is met, otherwise set to 0. N1 is the volume of a single voxel (fixed at 0.125 mm³); N1 is the total number of primes in the point cloud division.

[0072] S323: When the volume difference is greater than the preset threshold, it is determined that there is surface chipping damage.

[0073] For example, if the volume difference exceeds 50 mm³, it is determined that the power receiving shoe has surface chipping damage, and the maximum depth and projected area of ​​the chipping can be calculated.

[0074] This application's technical solution constructs a three-dimensional distance field and calculates the signed distance function (SDF) difference between the target point cloud and the standard template point cloud. The signed distance function (SDF) can accurately describe the distance and direction (inside / outside) from any point in space to the point cloud surface, thereby achieving sensitive capture of tiny defect areas. Then, based on the SDF difference, discretized integration is performed, and the difference volume of each voxel is accumulated after the three-dimensional space is meshed, ultimately obtaining a high-precision defect volume value. This method overcomes the limitation of direct point cloud comparison in quantifying void regions, achieving sub-millimeter-level precise quantization of chip damage.

[0075] In a specific illustrative embodiment, the contour inflection points of the electric receiving shoe include upper edge inflection points and lower edge inflection points. The calculation of geometric parameters and surface condition determination of the electric receiving shoe based on the fusion processing and recognition results includes a thickness calculation step. The thickness calculation step includes: S331: Based on the inflection point of the upper edge contour of the power receiving shoe, fit two spatial straight lines on both sides of the inflection point of the upper edge contour respectively. S332: Calculate the intersection of two spatial lines; S333: Based on the inflection point of the upper edge contour, draw a vertical straight line along the height direction of the power receiving shoe to obtain the first lower edge inflection point. Calculate the spatial distance between the intersection point and the first lower edge inflection point, and use the spatial distance as the thickness value of the power receiving shoe. The upper edge inflection point of the power receiving shoe refers to the transition point on the upper surface of the junction between the carbon sliding plate and the bracket on the contour curve of the power receiving shoe; the lower edge inflection point of the power receiving shoe refers to the lowest point of the bottom of the power receiving shoe on the contour curve. The lowest point of the bottom contour of the power receiving shoe is obtained by performing height direction extreme value search on the point cloud data, which is the lower edge inflection point.

[0076] like Figure 5 As shown, the blue lines represent the inflection points on the contour curve of the power receiving shoe. When calculating the thickness of the power receiving shoe, an upper edge inflection point is selected, and two spatial straight lines V1 are fitted on both sides of this upper edge inflection point. Figure 5 (The red line shown), spatial line V2 ( Figure 5 The green line shown); spatial line V1 ( Figure 5 The green line shown), the spatial line V2 ( Figure 5 The blue straight line shown intersects at the intersection point S1; from the intersection point S1, draw a vertical straight line along the height direction of the power receiving shoe to the lower edge of the power receiving shoe to obtain the first lower edge inflection point S2, calculate the spatial distance H between the intersection point S1 and the first lower edge inflection point S2, and the spatial distance value H is the thickness value of the power receiving shoe.

[0077] This application's technical solution, based on the precise positioning of the contour inflection point, fits spatial straight lines on both sides of the inflection point and calculates the intersection point. This intersection point corresponds to the theoretical transition point at the connection between the carbon slide plate's working surface and the bracket, serving as a reference point for thickness measurement and possessing clear geometric significance. Furthermore, the spatial Euclidean distance between this intersection point and the inflection point of the lower edge of the power receiving shoe (obtained through the extreme value of the point cloud height) is calculated. Since this distance is directly calculated based on three-dimensional spatial coordinates, it effectively avoids the projection error caused by the tilt of the power receiving shoe in traditional two-dimensional measurements, achieving accurate and stable measurement of the carbon slide plate's thickness.

[0078] In a specific illustrative embodiment, the three-dimensional point cloud data and the two-dimensional color image are fused and recognized. Based on the results of the fusion and recognition, the geometric parameters of the power receiving shoe are calculated and the surface condition is determined, including a height calculation step. The height calculation step includes: S341: Obtain the upper and lower edge inflection points of the power receiving shoe through image preprocessing; Specifically, an upper edge inflection point of the power receiving shoe is obtained through image preprocessing. A lower edge inflection point .

[0079] S342: Based on the known fixed installation angle of the camera, the calibration distance from the camera center to the rail surface, and the coordinates of the upper and lower edge inflection points, spatial trigonometric geometric relationships are calculated to obtain the vertical height value of the power receiving shoe.

[0080] The fixed mounting angle of the camera refers to the downward tilt angle of the camera module mounted on the brackets on both sides of the rail.

[0081] Specifically, such as Figure 6 As shown, the camera module is installed at a fixed angle. O is the origin of the coordinate system. Based on the upper edge inflection point P1 and the lower edge inflection point P2, ... and The coordinates of the direction can then be known. , , Given the lengths of the three sides, the solution can be found using these lengths. Angle value. By combining the installation position of the 3D line scan camera on site, the coordinate system of the camera module center can be obtained, and the included angle can be calculated. At the same time, the height of the lower edge of the power receiving shoe at a fixed position on a certain vehicle from the surface of the rail is measured manually. And calculate the lower edge inflection point of the corresponding three-dimensional model of the electric boot. Distance to the center point The angle between the line perpendicular to the ground and the line perpendicular to the ground This allows us to obtain the vertical distance component from the camera center to the horizontal plane containing the lower edge of the power receiving shoe. The distance from the camera center to the track surface is... When measuring the height of other power receiving boots, because... If the height remains constant, then the current height of the power receiving shoe above the rail surface is [value missing]. ,in, The lower edge inflection point of the power receiving shoe The projected length in the direction perpendicular to the track plane is... .

[0082] This application's technical solution transforms measurements in the camera coordinate system into absolute height relative to the rail surface. Using the camera's known fixed installation angle and the calibrated vertical distance from the camera center to the rail surface, combined with the coordinates of the upper and lower edge inflection points obtained through image preprocessing, a spatial triangular geometric relationship is constructed for calculation. This method fully utilizes the spatial geometric characteristics of 3D point clouds, converting the relative position of the power receiving shoe in the camera's field of view into its true vertical height relative to the rail surface. This eliminates measurement reference offsets caused by installation position fluctuations or train vibrations, achieving accurate height measurement under non-contact dynamic conditions.

[0083] In a specific illustrative embodiment, the three-dimensional point cloud data and the two-dimensional color image are fused and recognized. Based on the results of the fusion and recognition, the geometric parameters of the power receiving shoe are calculated and the surface condition is determined, including a length calculation step. The length calculation step includes: S351: Obtain the coordinates of the left endpoint and the right endpoint of the upper edge of the electric shoe through image preprocessing; Specifically, the upper left endpoint is calculated based on the image preprocessing algorithm. and the right end point .

[0084] S352: Calculate the spatial Euclidean distance between the left endpoint and the right endpoint of the upper edge, and use the spatial Euclidean distance as the length value of the power receiving shoe.

[0085] Specifically, the spatial Euclidean distance between the left endpoint and the right endpoint of the upper edge. The spatial Euclidean distance L is the length of the power receiving shoe.

[0086] This application's technical solution directly calculates the Euclidean distance between the left and right endpoints of the upper edge, obtained through image preprocessing, as the length of the electric receiving shoe. Because it uses 3D point cloud data, the calculated distance accurately reflects the actual extension length of the carbon sliding plate in 3D space, unaffected by minor rotations or tilts of the electric receiving shoe in the horizontal plane. Compared to the length measurement errors caused by perspective distortion in 2D images, this method achieves accurate and reliable measurement of the electric receiving shoe's length, providing a precise data foundation for wear trend analysis.

[0087] In a specific illustrative embodiment, the three-dimensional point cloud data and the two-dimensional color image are fused and recognized. Based on the results of the fusion and recognition, the geometric parameters of the power receiving shoe are calculated and the surface condition is determined, including a tilt angle calculation step. The tilt angle calculation step includes: S361: Obtain the horizontal reference direction in the camera coordinate system; S362: Obtain the coordinates of the left and right endpoints of the upper edge of the power receiving shoe through image preprocessing, and calculate the direction of the upper edge line of the power receiving shoe in the camera coordinate system; S363: Calculate the angle between the direction of the upper edge straight line and the horizontal reference direction, as the tilt angle of the power receiving shoe.

[0088] Specifically, given two points on the horizontal X-axis in the camera coordinate system... and The coordinates are obtained, and the left endpoint of the upper edge is obtained according to the image preprocessing algorithm. and the right end point The slope of two straight lines can be calculated based on the coordinates of four points. :

[0089]

[0090] The angle between two straight lines can be obtained using the slope formula. :

[0091] This application's technical solution uses the horizontal reference direction (such as the X-axis) in the camera coordinate system as a reference. It obtains the spatial direction vector of the straight line on the upper edge of the power receiving shoe through image preprocessing, and then calculates the angle between the two as the tilt angle. Since the 3D point cloud provides accurate spatial direction information, this angle directly reflects the true tilt attitude of the power receiving shoe relative to the horizontal plane. This method provides a quantitative basis for evaluating the installation status of the power receiving shoe, the swing amplitude during operation, and abnormal tilting (such as torsion spring failure), effectively compensating for the shortcomings of traditional two-dimensional detection in accurately measuring spatial angles.

[0092] The second aspect of this disclosure discloses an online monitoring system for power receiving boots, used to implement the online monitoring method for power receiving boots disclosed in the first aspect of this disclosure, such as... Figure 2 As shown, the online monitoring system for the power receiving shoe includes: A 3D line scan camera is used to acquire 3D point cloud data of the power receiving shoe; A two-dimensional color area array camera is used to acquire two-dimensional color images of the power receiving shoe; The upper-level fusion and recognition module connects a 3D line scan camera and a 2D color area scan camera to receive 3D point cloud data and 2D color images, and perform fusion processing and recognition. The control module, connecting the 3D line scan camera, the 2D color area scan camera, and the upper-level fusion and recognition module, is used for: The system monitors train travel information in real time, generates an initial trigger signal based on the train travel information, and controls a 3D line scan camera to collect 3D point cloud data of the power receiving shoe based on the initial trigger signal; the train travel information includes the train travel direction and the train travel speed. A secondary trigger signal is generated based on the 3D point cloud data. The 2D color area array camera is controlled to capture images of the power receiving shoe based on the secondary trigger signal, and a 2D color image of the power receiving shoe is acquired. The upper-level fusion and recognition module controls the fusion processing and recognition of three-dimensional point cloud data and two-dimensional color images, and calculates the geometric parameters of the power receiving shoe and judges its surface condition based on the results of the fusion processing and recognition.

[0093] In a specific illustrative embodiment, the electric boot monitoring system is installed at a non-continuous position of the three rails and includes an upper surface detection module, a lower surface detection module, a higher-level fusion and identification module, and a control module. Wherein, as... Figure 3As shown, the upper surface detection module includes a first upper surface detection group and a second upper surface detection group, deployed on the left and right sides of the rail, respectively. Each upper surface detection group includes one 3D line scan camera and one 2D color area scan camera. The lower surface detection module includes a first lower surface detection group and a second lower surface detection group, deployed on the left and right sides of the rail, respectively. Each lower surface detection group includes one 3D line scan camera and one 2D color area scan camera. Each detection module is installed at a 35° angle to the vertical baseline to ensure that the point cloud data between the upper and lower surface detection modules can be converted within the same coordinate system. The upper surface detection unit is used for acquiring images of the surface of the power receiving shoe's carbon sliding plate, the power receiving shoe's swing arm, and the torsion spring. The lower surface detection unit is used for acquiring images of the power receiving shoe's base bolts and the lower part of the swing arm. The system utilizes the upper surface detection module of the power receiving shoe to collect 3D point cloud data of the upper surface for high-precision 3D reconstruction, while simultaneously acquiring high-resolution 2D images. Similarly, the lower surface detection module collects high-precision point cloud data of the lower surface of the power receiving shoe, also acquiring high-resolution 2D images. These 3D point cloud data and high-resolution 2D images are then sent to the upper-level fusion and recognition module, which performs fusion processing and recognition based on the data. Specifically, when a train enters the detection area, the through-beam laser sensor is interrupted by the wheels or the vehicle body, generating a first trigger signal. The control module uses this first trigger signal to confirm the train's direction and activates the speed-measuring radar to collect the train's speed in real time. Based on a preset acquisition start position (usually the moment the power receiving shoe is about to enter the camera's field of view) and the current vehicle speed, the control module calculates the delay time and generates an initial trigger signal. This initial trigger signal is output to the 3D line scan camera in RS422 differential level format, triggering it to begin continuous line scan acquisition. The 3D line scan camera uses the principle of laser triangulation, outputting a contour line (containing 4096 points) with each scan. As the train moves, it accumulates 3D point cloud data covering the entire power receiving shoe. The point cloud data stream output by the 3D line scan camera in real time is sent to the buffer of the upper-level fusion and recognition module. The control module performs a simple height threshold judgment based on the point cloud data: when a sufficient number of point clouds appear in multiple consecutive contours within a preset depth of field, it is determined that the power receiving shoe has entered the center area of ​​the camera's field of view. At this time, the control module precisely delays the time according to the current train speed to ensure that when the power receiving shoe moves to the ideal shooting position of the 2D camera, it outputs a secondary trigger signal to the 2D color area scan camera. After receiving the secondary trigger signal, the 2D color area scan camera captures a high-definition color image (exemplarily, 5 megapixels) using global exposure and uploads the image to the upper-level fusion and recognition module.The control module controls the upper-level fusion and recognition module to perform iterative nearest point (ICP) registration on the 3D point cloud data to eliminate attitude deviations, extract contours and fit curves, accurately locate inflection points using the finite difference method and then perform straight line fitting. Based on the fitting results, the geometric parameters such as the thickness, height, length and tilt angle of the electric boot are accurately calculated. At the same time, the 2D color image is converted to the Lab color space to extract texture features for anomaly detection. Surface burning damage is identified through morphological processing and connected component analysis. Surface chipping defects are detected by comparing the volume difference between the registered point cloud and the standard template, thereby achieving deep fusion and comprehensive recognition of 3D geometric shape and 2D color texture information.

[0094] A third aspect of this disclosure provides a computer program product comprising computer program code that, when executed on a computer, causes the computer to perform the online monitoring method for the power receiving shoe described in the above embodiments. This computer program can be mounted on a train power receiving shoe, for example, a control unit of the power receiving shoe, for controlling the train power receiving shoe.

[0095] A fourth aspect of this disclosure provides a computer-readable storage medium storing program code that is executed by one or more processors. When the program code is executed on the processor, it causes an apparatus including one or more processors to perform the online monitoring method for the power receiving shoe described in the above embodiments. The processor running this computer-readable medium can be mounted on a train power receiving shoe for controlling the train power receiving shoe.

[0096] It should be understood that when the modules or units described herein are implemented using software, they can be implemented in whole or in part as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0097] A fifth aspect of this disclosure provides a chip system including a processor, or the chip system including a memory and a processor, for calling computer programs or computer instructions stored in the memory to cause the processor to execute the interactive methods involved in the above embodiments. The chip system can be a single chip or a chip module composed of multiple chips. The chip system can be mounted on a train power receiving shoe for controlling the power receiving shoe. Optionally, the memory of the control unit is an in-chip storage unit, such as a register, cache, etc. The storage unit can also be an external storage unit located in the wireless access device terminal, such as read-only memory (ROM) or other types of static storage devices that can store static information and instructions, random access memory (RAM), etc.

[0098] Those skilled in the art will recognize that the modules, units, and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0099] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for online monitoring of a power receiving shoe, characterized in that, include: The system monitors train travel information in real time, generates an initial trigger signal based on the train travel information, and controls a 3D line scan camera to collect 3D point cloud data of the power receiving shoe based on the initial trigger signal; the train travel information includes the train's direction of travel and train speed. A secondary trigger signal is generated based on the three-dimensional point cloud data, and a two-dimensional color area array camera is controlled based on the secondary trigger signal to capture the electric receiving shoe and acquire a two-dimensional color image of the electric receiving shoe. The three-dimensional point cloud data and the two-dimensional color image are fused and recognized. Based on the results of the fusion and recognition, the geometric parameters of the power receiving shoe are calculated and the surface condition is judged.

2. The online monitoring method for power receiving boots according to claim 1, characterized in that, The fusion processing and recognition of the three-dimensional point cloud data and the two-dimensional color image includes an image preprocessing step, which includes: The three-dimensional point cloud data is subjected to iterative nearest point (ICP) registration to obtain the first point cloud data aligned with the standard template point cloud; Extract the contour data of the power receiving boot from the first point cloud data; The contour data is fitted with a B-spline curve to obtain a contour curve function; Calculate the difference value of the contour curve function, and locate the contour inflection point of the electric receiving shoe based on the difference value; Two spatial straight lines are obtained by performing linear fitting on the data on both sides of the contour inflection point; The geometric parameters of the power receiving boot are calculated and the surface condition is determined based on the spatial straight line fitted by the straight line.

3. The online monitoring method for the power receiving shoe according to claim 1 or 2, characterized in that, The process involves fusing and recognizing the three-dimensional point cloud data and the two-dimensional color image, and calculating the geometric parameters and determining the surface condition of the power receiving shoe based on the results of the fusing and recognizing. This includes a surface burn determination step, which comprises: Locate the electric boot region in the two-dimensional color image and convert the electric boot region into the Lab color space; The texture features of the power-receiving shoe region are extracted in the Lab color space, and anomaly detection is performed. When an abnormal region is detected by the anomaly detection, image segmentation and morphological processing are performed on the abnormal region to obtain at least one connected component. Calculate the area and shape factor of each connected region. When the area of ​​the connected region is greater than or equal to a first area threshold and the shape factor is within a preset range, it is determined that the electric boot has surface burn damage.

4. The online monitoring method for the power receiving shoe according to claim 1 or 2, characterized in that, The process of fusing and recognizing the three-dimensional point cloud data and the two-dimensional color image, and calculating the geometric parameters and judging the surface condition of the power receiving shoe based on the results of the fusing and recognizing, includes a surface spalling judgment step, which includes: The three-dimensional point cloud data is registered and aligned with the standard template point cloud to obtain the first three-dimensional point cloud data; Calculate the volume difference between the first 3D point cloud data and the standard template point cloud; When the volume difference is greater than a preset threshold, it is determined that there is surface chipping damage.

5. The online monitoring method for the power receiving shoe according to claim 4, characterized in that, The calculation of the volume difference between the first 3D point cloud data and the standard template point cloud includes: Construct a three-dimensional distance field and calculate the difference of the signed distance function (SDF) between the target point cloud and the standard template point cloud; The volume difference is calculated by discretized integration based on the difference in the symbolic distance function (SDF).

6. The online monitoring method for the power receiving shoe according to claim 2, characterized in that, The contour inflection points of the power receiving boot include upper edge inflection points and lower edge inflection points. The calculation of geometric parameters and surface condition judgment of the power receiving boot based on the fusion processing and recognition results includes a thickness calculation step, which includes: Based on the inflection point of the upper edge contour of the power receiving shoe, two spatial straight lines are fitted on both sides of the inflection point of the upper edge contour. Calculate the intersection point of the two spatial lines; Based on the upper edge contour inflection point, draw a vertical straight line along the height direction of the power receiving shoe to the lower edge to obtain the first lower edge inflection point. Calculate the spatial distance between the intersection point and the first lower edge inflection point, and use the spatial distance as the thickness value of the power receiving shoe. The upper edge inflection point of the power receiving shoe refers to the transition point on the upper surface of the power receiving shoe at the junction of the carbon sliding plate and the bracket on the contour curve; the lower edge inflection point of the power receiving shoe refers to the lowest point of the bottom of the power receiving shoe on the contour curve.

7. The online monitoring method for the power receiving shoe according to claim 2, characterized in that, The process of fusing and recognizing the three-dimensional point cloud data and the two-dimensional color image, and then calculating the geometric parameters and determining the surface condition of the power receiving boot based on the results of the fusing and recognizing, includes a height calculation step. The height calculation step includes: The upper and lower edge inflection points of the power receiving shoe are obtained through image preprocessing; Based on the known fixed installation angle of the camera, the calibrated distance from the camera center to the rail surface, and the coordinates of the upper and lower edge inflection points, spatial trigonometric geometric relationships are calculated to obtain the vertical height value of the power receiving shoe.

8. The online monitoring method for the power receiving shoe according to claim 2, characterized in that, The process of fusing and recognizing the three-dimensional point cloud data and the two-dimensional color image, and calculating the geometric parameters and determining the surface condition of the power receiving boot based on the results of the fusing and recognizing, includes a length calculation step, which includes: The coordinates of the left endpoint of the upper edge and the right endpoint of the upper edge of the electric boot are obtained through image preprocessing. Calculate the spatial Euclidean distance between the left endpoint of the upper edge and the right endpoint of the upper edge, and use the spatial Euclidean distance as the length value of the power receiving shoe.

9. The online monitoring method for the power receiving shoe according to claim 1 or 2, characterized in that, The process of fusing and recognizing the three-dimensional point cloud data and the two-dimensional color image, and calculating the geometric parameters and determining the surface condition of the power receiving shoe based on the results of the fusing and recognizing, includes a tilt angle calculation step. The tilt angle calculation step includes: Obtain the horizontal reference direction in the camera coordinate system; The coordinates of the upper edge endpoint of the power receiving shoe are obtained through image preprocessing, and the direction of the upper edge line of the power receiving shoe in the camera coordinate system is calculated. Calculate the angle between the direction of the upper edge straight line and the horizontal reference direction, and use it as the tilt angle of the power receiving shoe.

10. An online monitoring system for power receiving boots, characterized in that, The method for implementing the online monitoring of the power receiving shoe according to any one of claims 1-9 includes: A 3D line scan camera is configured to acquire 3D point cloud data of the power receiving shoe; A two-dimensional color area array camera is configured to acquire two-dimensional color images of the power receiving shoe; The upper-level fusion and recognition module, connected to the 3D line scan camera and the 2D color area scan camera, is configured to receive the 3D point cloud data and the 2D color image, and perform fusion processing and recognition. The control module, connected to the 3D line scan camera, the 2D color area scan camera, and the upper-level fusion and recognition module, is configured as follows: The system monitors train travel information in real time, generates an initial trigger signal based on the train travel information, and controls the 3D line scan camera to collect 3D point cloud data of the power receiving shoe based on the initial trigger signal; the train travel information includes the train travel direction and the train travel speed. A secondary trigger signal is generated based on the three-dimensional point cloud data, and the two-dimensional color area array camera is controlled to capture the electric receiving shoe based on the secondary trigger signal to acquire a two-dimensional color image of the electric receiving shoe; The upper-level fusion and recognition module is controlled to perform fusion processing and recognition on the three-dimensional point cloud data and the two-dimensional color image, and to calculate the geometric parameters and determine the surface condition of the power receiving shoe based on the results of the fusion processing and recognition.

Images