Collector shoe carbon contact strip abrasion detection method and system based on single 3D camera

Abstract

The invention discloses a collector shoe carbon slide plate abrasion detection method based on a single 3D camera. The detection method comprises three parts of collector shoe three-dimensional image acquisition, carbon slide plate base plane establishment and collector shoe abrasion calculation. The invention further discloses a collector shoe carbon contact strip abrasion detection system based on the single 3D camera. According to the method, abrasion is measured through the single 3D camera, the manufacturing cost is low, calibration and splicing fusion work of double cameras does not need to be carried out, and the calculation complexity is reduced. According to the method, the thickness information of the surface of the whole carbon slide plate can be obtained, the abrasion error of each point on the surface of the carbon slide plate can be reduced by establishing the base plane, the measurement precision is improved, and the surface information can also be used for analyzing the eccentric abrasion of the carbon slide plate.

Description

Technical Field

[0001] The present invention belongs to the field of collector shoe carbon slide image detection, and relates to a collector shoe carbon slide wear detection method, and in particular to a collector shoe carbon slide wear detection method and system based on a single 3D camera. Background Art

[0002] The collector shoe carbon slide plate of a subway train is an important component of the current collecting device of an electrified railway, responsible for transferring current from the third rail to the train, thereby driving the train. The collector shoe carbon slide plate is widely used due to its good conductivity and wear resistance. However, as the operation time of subway trains increases, the collector shoe carbon slide plate will gradually wear out, and may have defects such as abnormal wear and defects; these defects will not only affect the normal operation of the train, but may also pose a threat to driving safety. Therefore, regular wear detection of the collector shoe carbon slide plate and timely detection and treatment of abnormal wear are of great significance to ensure the normal operation and driving safety of the subway.

[0003] At present, the traditional detection method of collector shoe carbon skateboard mainly relies on manual visual inspection and manual measurement, which has problems such as inconsistent measurement standards, large human errors, low measurement accuracy, and incomplete data.

[0004] There are also some machine vision-based detection methods for collector shoe carbon skateboards.

[0005] For example, a non-contact collector shoe carbon skateboard wear detection device and wear detection method (Chinese patent publication number: CN113865490A): two 3D cameras are respectively deployed above and below the collector shoe carbon skateboard to collect 3D information of the upper and lower surfaces of the collector shoe carbon skateboard, and the upper surface 3D information and the lower surface 3D information are fused into a 3D model of the collector shoe carbon skateboard. The thickness of the carbon skateboard is calculated according to the 3D model, and the wear of the carbon skateboard is determined; this method requires two sets of 3D cameras, which is relatively expensive, and the upper and lower sets of information need to be fused, and the calculation is relatively complicated.

[0006] For example, there is another method of inspecting the collector shoe carbon skateboard by using an area array camera (2D camera) to photograph the carbon skateboard from the side. By finding the upper and lower edges of the carbon skateboard, its thickness can be calculated. This method can only calculate the wear of the outer edge, and cannot obtain the wear of every point on the entire carbon skateboard surface.

[0007] In summary, it is necessary to study a fast, simple, robust and comprehensive method for detecting the wear of the carbon slide plate of the collector shoe. Summary of the invention

[0008] The purpose of the present invention is to propose a collector shoe carbon slide wear detection method based on a single 3D camera based on the three-dimensional information of the carbon slide, so as to achieve accurate measurement of the collector shoe wear.

[0009] The present invention also aims to provide a collector shoe carbon slide wear detection system based on a single 3D camera.

[0010] The technical solution of the present invention is as follows: A method for detecting wear of a collector shoe carbon slide plate based on a single 3D camera, comprising the following steps: S1. Collect a three-dimensional image of the upper surface of the collector shoe through a single 3D camera to obtain three-dimensional point cloud data of the collector shoe; S2, using the depth distance h from the collector shoe to the 3D camera to filter out the three-dimensional point cloud data within the range of h±50 mm, namely, including the three-dimensional point cloud data of the collector shoe carbon slide plate and the three-dimensional point cloud data of the collector shoe support arm; S3. Extract the three-dimensional point cloud coordinates of the collector shoe support arm according to the semantic segmentation algorithm; solve the three-dimensional point cloud coordinates of the carbon slide base plane according to the rotation axis and the rotation angle, and calculate the carbon slide base plane equation; S4, extracting the three-dimensional point cloud coordinates of the collector shoe carbon slide plate according to the depth threshold and semantic segmentation algorithm; The distance e, i.e., the thickness, from any point on the upper surface of the carbon slide plate to the base plane of the carbon slide plate is calculated, thereby obtaining the wear condition of the collector shoe carbon slide plate.

[0011] Furthermore, S3: Assume that the angle between the support arm plane and the carbon slide base plane is the rotation angle α, the straight line l that intersects the support arm plane and the carbon slide base plane is the rotation axis, substitute into the Rodriguez rotation formula, and solve the rotation matrix R that transforms the three-dimensional point cloud of the support arm plane to the three-dimensional point cloud of the carbon slide base plane; Substitute the rotation matrix R and the three-dimensional point coordinates p (x, y, z) on the arm plane into the formula , the three-dimensional point coordinates p' (x', y', z') on the carbon skateboard base plane can be obtained; Substituting p'(x',y',z') into the equation , the plane parameters A, B, C, and D are solved by the least square method to obtain the carbon skateboard base plane.

[0012] Furthermore, the S3: using a random sampling consensus method, the outlier points of the plane point cloud of the collector shoe support arm are removed.

[0013] Furthermore, the thickness: , wherein x'', y'', z'' are the three-dimensional coordinates of any point on the upper surface of the carbon slide; and A, B, C, D are the plane parameters of the carbon slide base plane.

[0014] Furthermore, the method also includes S5: first remove the three-dimensional point cloud data of the slope parts of the edges of the carbon skateboard on both sides to prevent interference with the final measurement results; include the calculated thickness e into a set, and then select the minimum value from the set to obtain the minimum remaining thickness of the carbon skateboard, and subtract the minimum remaining thickness of the carbon skateboard from the standard value of the initial thickness of the carbon skateboard to obtain the maximum wear of the collector shoe carbon skateboard.

[0015] Furthermore, in S1: the 3D camera is arranged beside the track and obliquely above the collector shoe, the vertical distance between the 3D camera and the collector shoe is 300-1000 mm, and the angle between the central axis of the 3D camera and the horizontal line is 30-70°.

[0016] Furthermore, in S1: in order to purify the three-dimensional point cloud data of the collector shoe, the three-dimensional point cloud data are filtered according to the discreteness and size characteristics of the interference points, so as to effectively remove the clutter and interference point cloud data.

[0017] Furthermore, in S1, the train passes through the speed measuring magnet, which emits a pulse of corresponding frequency, triggering the 3D camera to start collecting and taking pictures.

[0018] A collector shoe carbon plate wear detection system based on a single 3D camera, used to execute any one of the collector shoe carbon plate wear detection methods based on a single 3D camera, characterized in that the detection system includes a single 3D camera, a control unit and a data processing unit, the single 3D camera is arranged next to the track and located diagonally above the collector shoe, the control unit is communicatively connected to the 3D camera, and is used to complete the collection of collector shoe point cloud data.

[0019] Furthermore, the vertical distance between the 3D camera and the collector shoe is 300-1000 mm, and the angle between the central axis of the 3D camera and the horizontal line is 30-70°.

[0020] Compared with the prior art, the present invention has the following advantages: (1) Compared with the existing method that uses two sets of structured light vision modules installed on the upper and lower sides of the collector shoe carbon skateboard, the present invention uses a single 3D camera to measure wear, which has a lower manufacturing cost and does not require dual-camera calibration and splicing fusion, thereby reducing the complexity of the calculation.

[0021] (2) The method of the present invention can not only obtain the thickness information of the entire carbon slide surface, but also reduce the wear error of each point on the carbon slide surface by establishing a base plane, thereby improving the measurement accuracy. The surface information can also be used for the analysis of eccentric wear of the carbon slide.

[0022] In summary, the detection method of the present invention is a fast, simple, robust and comprehensive detection method for the wear of the collector shoe carbon slide. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of the collector shoe carbon slide wear detection method of the present invention; Figure 2 This is a schematic diagram of the installation of the 3D camera of the present invention; Figure 3 The 3D camera in Example 1 collects the image of the collector shoe carbon skateboard; Among them, 1-structured light vision module, 2-collector shoe support arm, 3-collector shoe carbon slide plate, 4-carbon slide plate base plane, 5-support arm plane; The depth distance from the surface of the h-3D camera to the collector shoe; α-the angle between the arm plane and the carbon slide base (i.e. the carbon slide base plane); l-The straight line where the support arm plane intersects the carbon skateboard base plane. DETAILED DESCRIPTION

[0024] In order to better understand the implementation process of the present invention, the specific implementation methods of the present invention will be described in detail below with reference to the accompanying drawings and implementation examples.

[0025] Example 1 This embodiment takes the scanning image of the train collector shoe carbon slide plate detection system as an example to demonstrate the process of measuring the wear of the collector shoe carbon slide plate 3.

[0026] Figure 1 It is the overall flow chart of the collector shoe carbon slide wear detection method based on a single 3D camera.

[0027] The collector shoe carbon slide wear detection method based on a single 3D camera in this embodiment includes three parts: collector shoe 3D image acquisition, collector shoe carbon slide base plane establishment, and collector shoe carbon slide wear calculation. Specifically: Step A: Acquisition of three-dimensional image of collector shoe.

[0028] Step A1: Arrange and install a collector shoe carbon plate wear detection system based on a single 3D camera, that is, a structured light vision module 1, such as Figure 2 As shown. The structured light vision module 1 is set as a single 3D camera (this embodiment uses a 3D camera from SICK), which is arranged and installed beside the track and obliquely above the collector shoe. The vertical distance between the 3D camera and the collector shoe is 300-1000mm, and the angle between the central axis of the 3D camera and the horizontal line is 30-70°. In this embodiment, the vertical distance between the 3D camera and the collector shoe is 400mm, and the angle between the central axis of the 3D camera and the horizontal line is 54°.

[0029] Step A2: When the train passes through the above detection system and the speed measuring magnet, the detection system emits a pulse of corresponding frequency, triggering the structured light vision module 1 installed above the collector shoe beside the track to take a photo, and collect a three-dimensional image of the upper surface of the collector shoe, and the brightness information is as follows: Figure 3 As shown, three-dimensional point cloud data is generated and uploaded to the analysis server.

[0030] Step A3: Based on the three-dimensional point cloud data acquired as above, the known depth distance h from the collector shoe to the 3D camera is used to filter out the three-dimensional point cloud data within the range of h±50mm. This part of the filtered three-dimensional point cloud data roughly covers the positions of the collector shoe carbon skateboard 3 and the collector shoe support arm 2, that is, the filtered three-dimensional point cloud data includes the three-dimensional point cloud data of the collector shoe carbon skateboard 3 and the collector shoe support arm 2 (the depth of field of the 3D camera is large, and other parts of the collector shoe (background) will be collected. The background can be quickly removed using the depth distance h).

[0031] Preferably, in order to purify the three-dimensional point cloud data of the collector shoe, these three-dimensional point cloud data are filtered according to the discreteness, size (such as length and area) and other characteristics of the interference points, so as to effectively remove the noise and interference point clouds. Step B: The collector shoe carbon slide plate 3 base plane is established.

[0033] Step B1: extract the collector shoe arm 2 according to the semantic segmentation algorithm, and obtain the three-dimensional point cloud coordinates of the collector shoe arm 2.

[0034] Specifically: The 3D camera outputs a two-dimensional grayscale image (light intensity information) and three-dimensional point cloud data at the same time, and these two types of data are one-to-one corresponding. Using a semantic segmentation algorithm (such as the unet network), the collector shoe arm 2 part on the two-dimensional grayscale image is labeled and trained, and then the collector shoe arm 2 can be detected and segmented. According to the above one-to-one correspondence principle, the three-dimensional point cloud data of the collector shoe arm 2 can be obtained; Preferably, the random sampling consensus method is used to remove the outliers of the point cloud of the arm plane 5. Ideally, the point cloud data collected by the 3D camera does not have these outliers, but there will be some noise interference in actual collection, which will cause loss of plane fitting accuracy if not removed.

[0035] Step B2: According to prior knowledge, there is a small angle α between the support arm plane 5 and the carbon slide base plane 4; and the straight line l intersecting the support arm plane 5 and the carbon slide base plane 4 is extracted and can be used as the rotation axis.

[0036] Substituting the above rotation axis and rotation angle α into the Rodriguez rotation formula (a well-known technique in the art), the rotation matrix R can be solved, that is, the rotation matrix R of the three-dimensional point cloud of the support arm plane 5 transformed to the three-dimensional point cloud of the new plane (carbon skateboard base plane 4).

[0037] By rotating the matrix R, the three-dimensional point cloud of the arm plane 5 is transformed into the three-dimensional point cloud of the new plane. At this time, the new plane is consistent with the carbon skateboard base plane 4. The solution formula is: ; Among them, p'(x', y', z') is the three-dimensional point coordinates of p(x, y, z) on the support arm plane 5 after rotation (i.e., the three-dimensional point coordinates on the new plane), i.e., the three-dimensional point coordinates on the carbon slide base plane 4; p (x, y, z) is the three-dimensional coordinate of the point on the arm plane 5; R is the rotation matrix R solved above.

[0038] Assume that the equation of the carbon skateboard base plane 4 is: , where x', y', z' are the p'(x', y', z') mentioned above. By solving the plane parameters A, B, C, D by the least squares method, the carbon skateboard base plane 4 can be solved.

[0039] Step C: Calculation of collector shoe carbon slide wear.

[0040] Step C1: extracting the three-dimensional point cloud data of the surface portion of the carbon slide plate according to the depth threshold and the semantic segmentation algorithm, that is, obtaining the three-dimensional point cloud data of the collector shoe carbon slide plate 3.

[0041] Calculate the distance e (thickness) from each point (point on the upper surface of the carbon slide) to the carbon slide base plane 4, the formula is , Among them, x'', y'', z'' are the three-dimensional coordinates of any point on the upper surface of the carbon slide; A, B, C, D are the plane parameters of the carbon slide base plane 4 obtained in the above step B2, which are known at this time.

[0042] Step C2: Based on prior knowledge, the point cloud data of the slope parts of the edges on both sides are eliminated to prevent interference with the final measurement results (the thickness of the slope parts at both ends of the carbon skateboard will be lower than that of the middle surface, which will affect the measurement of the lowest thickness point in the middle part. The slope parts at both ends cannot represent the actual wear of the carbon skateboard, so they need to be eliminated).

[0043] The remaining thickness values ​​are included in the set, and the minimum value is selected from the set to obtain the minimum remaining thickness of the carbon slide board. Finally, the maximum wear of the collector shoe carbon slide board can be obtained by subtracting the minimum remaining thickness of the carbon slide board from the initial thickness value of the carbon slide board (the thickness of the carbon slide board before use, which can be understood as the standard value here).

[0044] The collector shoe carbon skateboard wear detection method based on a single 3D camera of this embodiment is adopted to address the problems of complex calibration, high cost and incomplete detection results in the prior art. By utilizing the positional relationship between the collector shoe support arm 2 and the base of the collector shoe carbon skateboard 3, the base plane of the collector shoe carbon skateboard 3 is solved, and the distance from the surface point of the collector shoe carbon skateboard 3 to the base plane is calculated, thereby obtaining the wear of the collector shoe carbon skateboard 3.

[0045] The above content only presents an optimal implementation case proposed by the present invention, and its purpose is to elaborate on the core technical solution of the present invention, and it is by no means intended to limit its scope of application or possibility in any form. In other words, the technical framework and core concept of the present invention allow for a wide range of modifications and alternatives, and these flexible changes will not cause the essential characteristics of the relevant technical solutions to deviate from the basic spirit and coverage contained in the various embodiments of the present invention. Therefore, any reasonable adjustments and innovations based on this should be regarded as within the scope of protection of the present invention.

Claims

1. A method for detecting wear of a collector shoe carbon slide plate based on a single 3D camera, characterized in that: The method comprises the following: S1. Collect a three-dimensional image of the upper surface of the collector shoe through a single 3D camera to obtain three-dimensional point cloud data of the collector shoe; S2, using the depth distance h from the collector shoe to the 3D camera to filter out the three-dimensional point cloud data within the range of h±50 mm, namely, including the three-dimensional point cloud data of the collector shoe carbon slide plate and the three-dimensional point cloud data of the collector shoe support arm; S3. Extract the three-dimensional point cloud coordinates of the collector shoe support arm according to the semantic segmentation algorithm; solve the three-dimensional point cloud coordinates of the carbon slide base plane according to the rotation axis and the rotation angle, and calculate the carbon slide base plane equation; S4, extracting the three-dimensional point cloud coordinates of the collector shoe carbon slide plate according to the depth threshold and semantic segmentation algorithm; The distance e, i.e., the thickness, from any point on the upper surface of the carbon slide plate to the base plane of the carbon slide plate is calculated, thereby obtaining the wear condition of the collector shoe carbon slide plate.

2. A method for detecting wear of a collector shoe carbon slide plate based on a single 3D camera as claimed in claim 1, characterized in that: The S3: Assume that the angle between the support arm plane and the carbon slide base plane is the rotation angle α, the straight line l that intersects the support arm plane and the carbon slide base plane is the rotation axis, substitute into the Rodriguez rotation formula, and solve the rotation matrix R that transforms the three-dimensional point cloud of the support arm plane to the three-dimensional point cloud of the carbon slide base plane; Substitute the rotation matrix R and the three-dimensional point coordinates p (x, y, z) on the arm plane into the formula , the three-dimensional point coordinates p' (x', y', z') on the carbon skateboard base plane can be obtained; Substituting p'(x',y',z') into the equation , the plane parameters A, B, C, and D are solved by the least square method to obtain the carbon skateboard base plane.

3. A method for detecting wear of a collector shoe carbon slide plate based on a single 3D camera as claimed in claim 2, characterized in that: S3: using a random sampling consensus method to remove outliers in the plane point cloud of the collector shoe support arm.

4. A method for detecting wear of a collector shoe carbon slide plate based on a single 3D camera as claimed in claim 2, characterized in that: Said thickness: , wherein x'', y'', z'' are the three-dimensional coordinates of any point on the upper surface of the carbon slide; and A, B, C, D are the plane parameters of the carbon slide base plane.

5. A method for detecting wear of a collector shoe carbon slide plate based on a single 3D camera as claimed in any one of claims 1 to 4, characterized in that: The method also includes S5: first removing the three-dimensional point cloud data of the slope parts of the edges of the carbon slide board on both sides to prevent interference with the final measurement results; incorporating the calculated thickness e into a set, and then selecting the minimum value from the set to obtain the minimum remaining thickness of the carbon slide board, and subtracting the minimum remaining thickness of the carbon slide board from the standard value of the initial thickness of the carbon slide board to obtain the maximum wear of the collector shoe carbon slide board.

6. A method for detecting wear of a collector shoe carbon slide plate based on a single 3D camera as claimed in any one of claims 1 to 4, characterized in that: S1: The 3D camera is arranged beside the track and obliquely above the collector shoe, the vertical distance between the 3D camera and the collector shoe is 300-1000 mm, and the angle between the central axis of the 3D camera and the horizontal line is 30-70°.

7. A method for detecting wear of a collector shoe carbon slide plate based on a single 3D camera as claimed in any one of claims 1 to 4, characterized in that: S1: In order to purify the three-dimensional point cloud data of the collector shoe, the three-dimensional point cloud data are filtered according to the discreteness and size characteristics of the interference points, so as to effectively remove the noise and interference point cloud data.

8. A method for detecting wear of a collector shoe carbon slide plate based on a single 3D camera as claimed in any one of claims 1 to 4, characterized in that: S1: When a train passes through a speed measuring magnet, a pulse of a corresponding frequency is emitted, triggering the 3D camera to start collecting and taking pictures.

9. A collector shoe carbon plate wear detection system based on a single 3D camera, used to implement the collector shoe carbon plate wear detection method based on a single 3D camera as claimed in any one of claims 1 to 8, characterized in that: The detection system includes a single 3D camera, a control unit and a data processing unit. The single 3D camera is arranged beside the track and located diagonally above the collector shoe. The control unit is communicatively connected with the 3D camera to complete the collection of point cloud data of the collector shoe.

10. A collector shoe carbon slide wear detection system based on a single 3D camera as claimed in claim 9, characterized in that: The vertical distance between the 3D camera and the collector shoe is 300-1000 mm, and the angle between the central axis of the 3D camera and the horizontal line is 30-70°.

Images