Track anomaly detection equipment based on line scanning type point cloud processing method

Abstract

The invention relates to the field of intelligent detection, and provides track anomaly detection equipment based on a line scanning type point cloud processing method. The problems that in an existing track detection method, the manual work intensity is large, the accuracy is low, and front-back change comparison of the same position cannot be achieved are solved. The main scheme comprises a carrier, an identification reading device, a scanning device and a detection step. The carrier obtains the real-time running speed through the speed sensor, and the identification reading device can read the identification on the outer side of the track. A fan-shaped laser source on the scanning device vertically and downwards projects laser to form a laser scanning surface, and linear array type three-dimensional point cloud data of track surface contour information is obtained in real time. The detection steps comprise least square method straight line fitting, double-track edge point and middle position point identification, track key form and position parameter solving, sectional type detection and track abnormity and variation detection through a threshold value method and a reference template comparison method. The method is used for detecting shape and position abnormity and change of the track in real time, extra position and posture detection equipment is not needed, and key shape and position structure parameters of the track can be accurately determined.

Description

Technical Field

[0001] The invention belongs to the field of intelligent detection, and in particular relates to a method for scanning and detecting track shape and position anomalies and changes by using three-dimensional point cloud processing technology. Background Art

[0002] The long-term use of tracks will inevitably be accompanied by wear and tear, defects and deterioration. Abnormal track shape and position will pose a huge threat to train operation. Common and harmful shape and position abnormalities mainly include vertical unevenness of single track, abnormal width, distortion, concave and convex defects, abnormal horizontal height between double tracks, abnormal track gauge, etc. Manual regular inspection of track shape and position is labor-intensive, low in accuracy, and it is difficult to compare the changes before and after the same position. Summary of the invention

[0003] In order to achieve efficient and accurate real-time detection, the present invention provides a device for detecting track shape and position anomalies. The 3D point cloud processing technology is used to scan and detect the track's morphological features and abnormal changes. The location, type and degree of track structural anomalies can be determined in real time without the need for posture detection equipment such as a level, which greatly facilitates track maintenance and inspection. Different from other track detection methods such as image processing, it has the advantage of being more microscopic and accurate.

[0004] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0005] The present invention provides a track anomaly detection device based on a line scanning point cloud processing method, comprising:

[0006] Vehicle: The vehicle obtains the real-time running speed of the vehicle through a speed sensor; a mark reading device is fixed on the outside of the vehicle, and the device can read the marks arranged at equal intervals on the outside of the track;

[0007] Scanning device: The fan-shaped laser source on the scanning device projects a high-brightness laser vertically downward to form a laser scanning surface. The scanning device acquires linear array 3D point cloud data representing the track surface profile information in real time, that is, the y and z coordinates of the dense points distributed on the "one" laser line along the track width direction are acquired, and the x coordinate of the point cloud is obtained by combining the real-time speed of the vehicle and the timestamp of each frame of linear array point cloud data;

[0008] The detection steps are:

[0009] Step (1), perform least square method zy straight line fitting based on the linear array point cloud data to obtain the fitting straight line l: Ay+Bz+C=0;

[0010] Step (2), identifying the edge points of the two tracks and the midpoints of each track from each frame of the linear point cloud;

[0011] Step (3), obtaining key track geometry parameters based on the edge points and midpoints of the double tracks;

[0012] Step (4), reading the identification by the reading device to realize segmented detection;

[0013] Step (5) detects track shape and position anomalies and changes through two modes: threshold method and reference template comparison method.

[0014] In the above scheme, in step (1), the part of each frame of linear point cloud outside the sleeper area is taken as the point set, and the zy straight line fitting based on the least squares method is performed to determine the representation of the ground reference in the point cloud coordinate system.

[0015] In the above scheme, in step (2), the dual-track edge points are identified from each frame of the linear point cloud according to the zy change rate characteristics of each point, specifically:

[0016] Edge points refer to the dense linear point cloud P in each frame. i {(x i ,y i1 , z i1 ), (x i ,y i2 , z i2 )…,(x i ,y in , z in )}, the coordinate point closest to the edge of the double track, for the left edge point p of track α and track β i,αL and p i，βL , identify p by threshold judgment i，αL and p i，βL , that is, for the linear point cloud P i A point p on i，j (x i ,y i，j , z i,j ),

[0017] (z i，j -z i，j-1 ) / (y i，j -y i，j-1 )<-T R , j = 2, 3, 4, ..., n Formula 1

[0018] If equation 1 is satisfied, then p i，j is the left edge point, where T R Represents the change rate threshold. For the two identified left edge points, the one with the smaller y coordinate corresponds to the left edge point p of the left track α. i,αL , the larger one corresponds to the left edge point p of the right track β i，βL ;

[0019] For the right edge point p of track α and track β i,αR and p i，βRIdentify by:

[0020] (z i，j+1 -z i，j ) / (y i，j+1 -y i，j )>T R , j = 1, 2, 3, ..., n-1 Formula 2

[0021] If equation 2 is satisfied, then p i,j is the right edge point. For the two identified right edge points, the one with the smaller y coordinate corresponds to the right edge point p of the left track α. i，αR , the larger one corresponds to the right edge point p of the right track β i，βR ;

[0022] Then we can get the first frame of linear point cloud data P i The left and right edge points of the upper double track i，αL (x i ,y i，αL , z i，αL ), p i，αR (x i ,y i，αR , z i,αR ), p i，βL (x i ,y i，βL , z i，βL ) and p i，βR (x i ,y i，βR , z i，βR ).

[0023] In the above scheme, in step (2), the left and right edge points p of the double track are i，αL (x i ,y i，αL , z i，αL ), p i，αR (x i ,y i,αR , z i,αR ), p i，βL (x i ,y i，βL , z i，βL ) and p i，βR (x i ,y i，βR , z i，βR ) to determine the midpoint p that can represent the center of the track cross section i，αM (x i ,y i，αM , z i,αM ) and p i，βM (x i ,y i，βM , z i，βM ), where pi,αM is the point p between the left and right edges of track α i,αL and p i,αR The coordinate point between the median and p i，βM is the point p between the left and right edges of track β i，βL and p i，βR The coordinate point in between.

[0024] In the above scheme, in step (3), the key shape and position features of the track are obtained by line scanning, which specifically includes:

[0025] The width of the left and right rails is obtained by calculating the distance between the left and right edge points of each rail in the vz plane, and obtaining the width w of the rail α i，α , rail width w i，β ;

[0026] Track gauge: The track gauge W is obtained by calculating the distance between the right edge point of the left track and the left edge point of the right track in the yz plane. i ;

[0027] The height of each left and right track: The distance from the median coordinate point of each track to the ground reference fitting line in the yz plane is obtained to obtain the track α height h i，α , rail β height h i，β ;

[0028] Roll angle: The angle between the midpoints of the left and right rails and the ground reference fitting straight line in the yz plane is obtained to obtain the roll angle γ i ;

[0029] The smoothness of the cross-sectional end surface of the left and right rails: a window is defined with the midpoint of each rail as the center, and the zv fitting line of the coordinate points in the window is calculated. Then, the discreteness of these coordinate points relative to the fitting line is calculated to obtain the cross-sectional end surface smoothness of rail α. Rail β cross-section end surface smoothness

[0030] Pitch angle of left and right rails: Calculate the arctangent of the height change of the median coordinate point and the x difference between the two frames before and after to obtain the pitch angle θ of rail α i,α , track β track pitch angle θ i，β ;

[0031] Left and right track smoothness: Calculate the zx fitting line with the current frame and the previous frames as the regression point set, and then calculate the discreteness of these coordinate points relative to the fitting line to obtain the track α smoothness and track β track smoothness

[0032] In the above scheme, in step (4), segmented detection is performed by marking to facilitate the determination of the location where the abnormality occurs, specifically:

[0033] The identification reading device fixed on the carrier reads the identification corresponding serial number. The identification is arranged at equal intervals of Δx on the outside of the track. It is assumed that the abnormal frame is identified between the reading time of identification I and identification II, and from t I The calculation starts when the marker I is read at time t. The mth frame linear array data has parameter anomalies, i.e., the anomaly recognition time t m Corresponding to the mth frame of linear point cloud data after the identification I is read, the abnormal position can be accurately located according to the speed information obtained by the speed sensor:

[0034]

[0035] Where v m represents the real-time speed of the vehicle when acquiring the mth frame of data, t1 and v1 represent the time of the first frame from reading tag I and the real-time speed of the vehicle, respectively, m Indicates the distance from the abnormal point to the same scanning surface as marker I, x m_I It represents the distance from the outlier point to the marker I, v i Indicates the real-time speed of the vehicle when the i-th frame of linear point cloud data is obtained starting from reading to identification I.

[0036] In the above scheme, in step (5), track anomalies and changes are detected by combining the two modes A and B:

[0037] Mode A: Threshold discrimination method, that is, if a key shape and position characteristic parameter of a track exceeds the normal threshold range, it can be considered that a track shape and position abnormality occurs at the corresponding position of the current scanning frame;

[0038] Mode B: Benchmark template comparison method. Specifically, first complete a preliminary scan of all sections in the entire detection domain. According to the calculation method of the key shape and position feature parameters of the track, obtain the shape and position parameter vectors of each frame in each section to construct a benchmark template. Then, during the actual detection, use the interpolation method to calculate the difference vector between the shape and position parameter vector of the current detection frame and the benchmark template. Based on the difference vector, the size of the change in the corresponding features of each vector element compared to the initial benchmark can be determined, thereby judging the changes in the track shape and position on the time scale.

[0039] In the above scheme, the reference template comparison method specifically includes:

[0040] Taking the segment divided by markers I and II as an example, the 1st to nth frames of point cloud data are collected in this segment, and the shape and position parameter vectors S1, S2, ... S corresponding to each of these n frames of data are obtained. i , ... S n , forming a reference template, where any vector

[0041]

[0042] Among them, the first element x of the vector i From the x in formula 3 m_I The solution method is determined, which represents the distance from the i-th frame data to the identifier I, and the remaining elements are all the track shape and position parameters determined according to Table 1;

[0043] In actual detection, when the vehicle passes through the section divided by identifiers I and II, the parameter vectors corresponding to all frame linear point cloud data in the section are obtained, where a frame vector S m It is expressed as:

[0044]

[0045] If x i <x m <x i+1 , that is, S m The corresponding scanning position is between the reference template vector S i and S i+1 The difference vector ΔS is calculated by interpolation. m :

[0046]

[0047] ΔS m The difference between the actual detection and the initial benchmark is characterized, and the change in each track structure parameter is determined by its 2nd to 13th elements;

[0048] For example, in actual detection, the difference vector ΔS corresponding to the mth frame of a certain scanning position m The fourth element ΔW m If the absolute value is greater than the specified value, it means that the x after the mark I at the corresponding scanning position is m_I The track gauge at has a significant change compared to the initial reference. A positive value means that the track gauge has become larger, and a negative value means that the track gauge has become smaller.

[0049] The seventh element Δγ m If a positive or negative change greater than the specified value occurs, it means that the roll angle at the corresponding position has changed significantly compared to the pre-scan. If the vehicle still passes through this place at the same speed, there is a risk of overturning, and the place needs to be recalculated. Therefore, mode B can accurately obtain the location, type and degree of the change, which is convenient for targeted maintenance and inspection.

[0050] Beneficial effects:

[0051] The present invention is aimed at track detection. Based on the three-dimensional point cloud processing method, a unique line scanning detection equipment is designed to identify track morphological anomalies and changes. The track edge coordinate points are identified from the linear array point cloud data obtained by scanning, and the key shape and position structural features of the track are calculated accordingly.

[0052] A method for determining the ground reference based on linear point cloud was designed using the least squares method, without the need to deploy additional posture detection sensing equipment such as a level; a marked segmented detection mechanism was designed to facilitate the location of abnormal positions; two detection modes, threshold discrimination method and benchmark comparison method, were designed to respectively identify anomalies and changes in track structure parameters.

[0053] Compared with existing intelligent track detection equipment, the present invention has a simple layout, low computational complexity, good real-time performance, and high detection efficiency. It can accurately determine the key shape and position structural parameters of the track, and can not only identify structural parameter anomalies through the threshold method, but also accurately obtain the shape and position structural changes. It can also accurately locate the location where the abnormal change occurs, which is convenient for repair and maintenance. In addition, compared with the conventional image processing detection method based on pixel RGB color values, the present invention can avoid the influence of lighting conditions and color differences, and is expected to achieve more reliable detection in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a structural schematic diagram of the present invention;

[0055] Figure 2 It is a diagram illustrating the ground reference fitting solution of the present invention;

[0056] Figure 3 It is a schematic diagram of the track edge point extraction principle of the present invention;

[0057] Figure 4 Table 1 is a diagram showing the calculation of key geometry parameters of the track according to the present invention;

[0058] Figure 5 It is a diagram illustrating the marked segmented detection mechanism of the present invention.

[0059] Description of Figure Numbers:

[0060] 1- scanning device, 2- carrier, 201- speed sensor, 202- identification reading device, 301- identification I, 302- identification II. DETAILED DESCRIPTION

[0061] like Figure 1 As shown, the present invention proposes a track anomaly detection equipment based on a line scanning point cloud processing method, which mainly includes a scanning device 1 and a carrier 2.

[0062] The scanning device 1 is installed at the front end of the vehicle 2. The scanning device can be a conventional point cloud data acquisition device such as a laser scanner. The vehicle can be a special track inspection vehicle or an actual running train. The fan-shaped laser source installed on the scanning device 1 projects the laser vertically downward to form a laser scanning surface. The scanning device 1 obtains the y and z coordinates of the dense points distributed on the "I" laser contour line along the width direction of the track, and their x coordinates are obtained by combining the real-time speed of the vehicle and the timestamp of each frame of linear point cloud data (the scanning device coordinate system is defined as: the vehicle running direction is x, the track width direction is v, and the vehicle vertical direction is z). That is, the linear array three-dimensional point cloud data {P1{(x1, y 1，1 , z 1，1 ), (x1, y 1，2 , z 1，2 )…,(x1,y 1n , z 1，n )}, P2{(x2, y 2，1 , z 2，1 ), (x2, y 2，2 , z 2，2 )…,(x2,y 2，n , z 2，n )}, …P i {(x i ,y i，1 , z i，1 ), (x i ,y i，2 , z i，2 )…,(x i ,y i，n , z i，n The vehicle speed is collected by a conventional speed sensor 201, which mainly obtains the wheel linear speed or rotational speed through various principles such as photoelectric, magnetoelectric, and Hall, and then obtains the vehicle running speed.

[0063] Based on the line scan data acquisition mode, a ground reference acquisition method is designed, such as Figure 2 As shown in the figure, the part of each frame of linear point cloud data outside the sleeper area is taken as the point set, and the least square method zy straight line fitting is performed to obtain the fitting line l: Ay+Bz+C=0. In this way, the representation of the ground reference in the point cloud coordinate system can be determined without the aid of a level meter, which is convenient for the subsequent solution of track parameters such as rail surface height and roll angle.

[0064] Based on the line scan data acquisition mode, an edge point extraction method is designed to extract the dual-track edge points from each frame of linear point cloud data in real time to accurately identify the left and right edges of the dual tracks. i {(x i ,yi1 , z i1 ), (x i ,y i2 , z i2 )…,(x i ,y in , z in )} respectively find the coordinate points closest to the edge of the double track. Figure 3 As shown, according to the zy change rate characteristics of each point, from the first frame linear point cloud P i Specifically, for track α and track , the left edge point p i,αL and p i,βL Due to the large height difference, the change rate of zy at this point will be very large, and the p can be effectively identified by threshold judgment. i,αL and p i，βL That is, for the linear point cloud P i A point p on i,j (x i ,y i，j , z i，j ),if

[0065] (z i，j -z i，h-1 ) / (y i，j -y i，h-1 )<-T R , j = 2, 3, 4, ..., n Formula 1

[0066] Then p i,j is the left edge point. R Indicates the change rate threshold. For the two identified left edge points, the one with the smaller y coordinate corresponds to the left edge point p of the left track α. i，αL , the larger one corresponds to the left edge point p of the right track β i，βL Similarly, for the right edge point p of track α and track β i,αR and p i，βR Identify by:

[0067] (z i,j+1 -z i,j ) / (y i,j+1 -y i,j )>T R ,j=1,2,3,...,n-1 Formula 2

[0068] In determining the i-th frame linear point cloud data P i The left and right edge points of the upper double track i，αL (x i ,y i，αL , z i，αL ), p i，αR (x i ,yi，αR , z i，αR ), p i，βL (x i ,y i，β: , z i，βL ) and p i，βR (x i ,y i，βR , z i，βR ) after that. The midpoint p that can represent the center of the track cross section can be determined i，αM (x i ,y i，αM , z i，αM ) and p i，βM (x i ,y i，βM , z i，βM ),in, i,αM is the point β between the left and right edges of track α i,αL and p i,αR The coordinate point between the median and p i,βM is the point p between the left and right edges of track β i，βL and p i,βR The coordinate point in between.

[0069] Furthermore, based on the calculation method of key track parameters shown in Table 1, the key track shape and position characteristics can be obtained by line scanning: track width w i，a , rail width w i，β 、Track gauge W i 、rail α height h i,α , rail β height h i，β , roll angle γ i 、Smoothness of rail α cross section end surface Rail β cross-section end surface smoothness Track α Track pitch angle θ i,α , track β track pitch angle θ i，β , rail α rail smoothness and track β track smoothness Specifically:

[0070] Width of left and right rails w i,α and w i，β , respectively obtained from the distances between the left and right edge points of each track in the yz plane:

[0071] According to the left edge point p of track α i，αL (x i ,y i，αL , z i，αL ) and the right edge point p i，αR (x i ,y i，αR , z i，αR )Sure:

[0072]

[0073] According to the left edge point p of track β i，βL (x i ,y i，βL , z i，βL ) and the right edge point p i，βR (x i ,y i，βR , z i，βR )Sure:

[0074]

[0075] Track gauge W i :The distance between the right edge point of the left track and the left edge point of the right track in the yz plane is obtained, according to the right edge point p of track α i,αR (x i ,y i,αR , z i，αR ) and point p on the left edge of track β i，βL (x i， y i，βL , z i，βL )Sure:

[0076]

[0077] Left and right rail height h i,α and h i,β :The distance from the median coordinate point of each track to the ground reference fitting straight line in the yz plane is obtained, according to the median point p of track α i，αM (x i ,y i,αM , z i,αM ) to the ground reference fitting straight line l: Ay+Bz+C=0 is determined by:

[0078]

[0079] Roll angle γ i : Calculate the angle between the midpoint line of the left and right rails and the ground reference fitting straight line in the yz plane to calculate p i，αM (x i ,y i,αM , z i,αM ) and p i，βM (x i ,y i，βM , z i，βM )’s slope:

[0080]

[0081] The roll angle

[0082] k' is the slope of the ground reference fitting straight line l: Ay+Bz+C=0, that is, k'=-A / B

[0083] Left and right rail cross-sectional end surface smoothness and A window is defined with the midpoint of each track as the center, and the zy fitting line of the coordinate points in the window is calculated. Then, the discreteness of these coordinate points relative to the fitting line is calculated.

[0084] The midpoint p of orbital α i，αM A window is defined as the center, and the coordinate point p on the upper end surface of track α in the window is calculated i，j ~p i，j+m The zy fitting straight line l z-y ,;

[0085] Calculate the distance from each point to l z-y , the distance d i，j ~d i，j+m ;

[0086] The standard deviation The larger the value, the worse the smoothness.

[0087] The midpoint of orbital β, p i，βM A window is defined as the center, and the coordinate point p on the upper end surface of the rail β in the window is calculated i，j ~p i，j+m The zy fitting straight line l z-y ;

[0088] Calculate the distance from each point to l z-y The distance d i，j ~d i，j+m ;

[0089] The standard deviation The larger the value, the worse the smoothness.

[0090] Left and right rail pitch angle θ i，α and θ i，β : Calculate the arc tangent value of the height change and x difference of the median coordinate point between the previous and next two frames;

[0091] According to the median point p corresponding to the track α on the i-th frame linear point cloud data i，αM (x i ,y i，αM , z i，αM ) Determine the height h of rail α i，α (See above i,α Solve), also based on the median point p in the previous frame of linear point cloud data i-1，αM (x i-1 ,y i-1，αM , z i-1，αM ) corresponds to the height h of the track α i-1，α ;

[0092] Then the longitudinal pitch angle of track α is:

[0093]

[0094] Do the same for track β.

[0095]

[0096] Left and right rail smoothness and The zx fitting line is calculated using the points in the current frame and the previous frames as the regression point set, and then the discreteness of these coordinate points relative to the fitting line is calculated;

[0097] The point p in the track α corresponding to the point cloud data of the i-th frame and several frames before it i-m，αM (x i ,y i-m，αM , z i-m，αM )~p i，αM (x i ,y i,αM , z i，αM ) is the regression point set, calculate the zx fitting line l z-x ;

[0098] Calculate the distance from each point to l z-x The distance d i-m ~d i ;

[0099] The standard deviation The larger the value, the worse the track smoothness;

[0100] Do the same for track β:

[0101] Furthermore, in order to facilitate the determination of the abnormality location, the present invention also proposes a segmented marking detection mechanism. Figure 4 As shown, markers I, II, III, etc. are arranged at equal intervals of Δx on the outside of the track. The markers can be RFID electronic tags, bar codes, QR codes, color codes, etc. A marker reading device is fixed on the train. When the device faces the marker, the corresponding serial number of the marker can be read. When the reading device reads a marker (such as Figure 4 When the mark I is shown, the line scanning surface at that moment is the same position scanning surface corresponding to the mark (such as the same position scanning surface of the mark I shown in the figure). Since the distance from the mark reading device to the scanning surface is x p , so the distance from the marker to its corresponding co-located scanning surface is also x p According to this position relationship, it is easy to find the specific location where the track anomaly occurs. For example, if the anomaly is identified at time t m, and it is at the time of reading the mark I and reading the mark II (t I and t II ), then we know that the anomaly must appear between the co-location scanning surface of marker I and the co-location scanning surface of marker II. Further, assuming that from t I At the moment when the marker I is read, the calculation starts. The mth frame of linear array data has parameter anomalies, that is, the anomaly recognition time t m Corresponding to the mth frame of linear point cloud data after the identification I is read. The abnormal position can be accurately located according to the speed information obtained by the speed sensor:

[0102]

[0103] In the formula, x m Indicates the distance from the abnormal point to the same scanning surface as marker I, x m_I It represents the distance from the outlier point to the marker I, v i Indicates the real-time speed of the vehicle collected by the speed sensor when acquiring the i-th frame (calculated from reading to identification I) of the linear array point cloud data.

[0104] In this way, by using the segmentation mechanism, the small segment Δx is used as the abnormality analysis unit, and the location where the abnormality occurs can be located more accurately.

[0105] Furthermore, according to the key track parameters corresponding to each frame of the linear point cloud determined in Table 1 and the segmented detection mechanism, the detection of track anomalies and changes is achieved through mode A and mode B:

[0106] Mode A: Use a simple threshold method to identify track anomalies. If the track width, track gauge, rail height, roll angle, and track pitch angle exceed the normal threshold range, it may affect the safe operation of the train and require maintenance. If the cross-sectional smoothness and track smoothness exceed the normal threshold range, the upper end surface of the rail may be damaged or defective, thus affecting the smoothness of the train operation and requiring repair.

[0107] Mode B: To accurately and effectively judge abnormal track changes, the track shape and position changes are determined by comparing with the reference template. Specifically, first let the vehicle run slowly and evenly to complete the pre-scan of the entire detection section, and obtain the track shape and position parameters corresponding to all linear point cloud data in the pre-scan process according to Table 1. Figure 4 Taking the section divided by the markers I and II as an example, the 1st to nth frames of point cloud data are collected in this section, and the shape and position parameter vectors S1, S2, ... S corresponding to each of the n frames of data are obtained. i , ... S n , forming a reference template, where any vector

[0108]

[0109] Among them, the first element x of the vector i From formula (3), x m_I The solution method is determined, and the remaining elements are all the orbital shape and position parameters determined according to Table 1.

[0110] In actual detection, when the vehicle passes through the section divided by markers I and II, the parameter vectors corresponding to all frame line array point cloud data in the section will also be obtained. m For example, it means

[0111]

[0112] If x i <x m <x i+1 , that is, S m The corresponding scanning position is between the reference template vector S i and S i+1 The difference vector ΔS can be calculated by interpolation. m :

[0113]

[0114] ΔS n It represents the difference between the actual detection and the initial reference. The change of each track structure parameter can be judged by its 2nd to 13th elements. For example, in actual detection, the difference vector ΔS corresponding to a certain scanning position (taking the mth frame as an example) m The fourth element ΔW m The absolute value is larger, which means that the corresponding scanning position (x after the mark I) m_I The track gauge at the location of the track has changed significantly compared to the initial reference. If the value is positive, it means that the track gauge has become larger, otherwise it has become smaller. For example, the seventh element Δγ m If a large positive or negative change occurs, it means that the roll angle at the corresponding position has changed significantly compared to the pre-scan. If the vehicle still passes through this place at the same speed, there may be a risk of overturning, and the location needs to be recalculated. Therefore, through mode B, the location, type and degree of the change can be accurately obtained, which is convenient for targeted maintenance and repair.

[0115] The above track anomaly detection equipment based on the line scanning point cloud processing method can selectively or simultaneously work in mode A and mode B: mode A is used to detect shape and position structural anomalies, while mode B is used to detect front and back anomalies.

Claims

1. A track anomaly detection device based on a line scanning point cloud processing method, characterized in that: include: Vehicle: The vehicle (2) obtains the real-time running speed of the vehicle through a speed sensor (201); a mark reading device (202) is fixedly connected to the outside of the vehicle, and the device can read marks arranged at equal intervals on the outside of the track; Scanning device (1): The fan-shaped laser source on the scanning device (1) projects a high-brightness laser vertically downward to form a laser scanning surface. The scanning device (1) acquires linear array three-dimensional point cloud data representing the contour information of the track surface in real time, that is, the y and z coordinates of the dense points distributed on the "single" laser line along the width direction of the track are acquired, and the x coordinate of the point cloud is obtained by combining the real-time speed of the vehicle and the timestamp of each frame of linear array point cloud data; The detection steps are: Step (1), perform least square method zy straight line fitting based on the linear array point cloud data to obtain the fitting straight line l: Ay+Bz+C=0; Step (2), identifying the edge points of the two tracks and the midpoints of each track from each frame of the linear point cloud; Step (3), obtaining key track geometry parameters based on the edge points and midpoints of the double tracks; Step (4), reading the identification by the reading device (202) to realize segmented detection; Step (5) detects track shape and position anomalies and changes through two modes: threshold method and reference template comparison method.

2. The track anomaly detection device according to claim 1, characterized in that: In step (1), the part of each frame of the linear point cloud outside the sleeper area is taken as the point set, and the zy straight line fitting based on the least squares method is performed to determine the representation of the ground reference in the point cloud coordinate system.

3. The track anomaly detection device according to claim 1, characterized in that: In step (2), the dual-track edge points are identified from each frame of the linear point cloud according to the zy change rate characteristics of each point, specifically: Edge points refer to the dense linear point cloud P in each frame. i {(x i ,y i1 , z i1 ), (x i ,y i2 , z i2 )…,(x i ,y in , z in )}, the coordinate point closest to the edge of the double track, for the left edge point p of track α and track β i,αL and p i,βL , identify p by threshold judgment i，αL and p i，βL , that is, for the linear point cloud P i A point p on i，j (x i ,y i，j , z i，j ), (z i，j -z i，j-1 ) / (y i，j -y i，j-1 ) < -T R , j = 2, 3, 4, ..., n Equation 1 If equation 1 is satisfied, then p i，j is the left edge point, where T R Represents the change rate threshold. For the two identified left edge points, the one with the smaller y coordinate corresponds to the left edge point p of the left track α. i，αL , the larger one corresponds to the left edge point p of the right track β i，βL ; For the right edge point p of track α and track β i,αR and p i，βR Identify by: (z i,j+1 -z i,j ) / (y i,j+1 -y i,j )>T R , j = 1, 2, 3, ..., n - 1 Equation 2 If equation 2 is satisfied, then p i，j is the right edge point. For the two identified right edge points, the one with the smaller y coordinate corresponds to the right edge point p of the left track α. i,αR , the larger one corresponds to the right edge point p of the right track β i,βR ; The i-th frame linear point cloud data P can be obtained i The left and right edge points of the upper double track p i，αL (x i ,y i，αL , z i，αL ), p i，αR (x i ,y i，αR , z i，αR ), p i，βL (x i ,y i，βL , z i，βL ) and p i，βR (x,y i，βR , z i，βR ).

4. The track anomaly detection device according to claim 3, characterized in that: In step (2), through the left and right edge points p of the double track i，αL (x i ,y i，αL , ，αL ), p i，αR (x i ,y i，αR , z i，αR ), p i，βL (x i ,y i，βL , z i，βL ) and p i，βR (x i ,y i，βR , z i，βR ) to determine the midpoint p that can represent the center of the track cross section i，αM (x i ,y i，αM , z i，αM ) and p i，βM (x i ,y i，βM , z i，βM ), where p i，αM is the point p between the left and right edges of track α i,αL and p i,αR The coordinate point between the median and p i，βM is the point p between the left and right edges of track β i，βL and p i，βR The coordinate point in between.

5. The track anomaly detection device according to any one of claims 1 or 4, characterized in that: In step (3), the key shape and position features of the track are obtained by line scanning, which specifically includes: The width of the left and right rails is obtained by calculating the distance between the left and right edge points of each rail in the yz plane to obtain the width w of the rail α i，α , rail width w i，β ; Track gauge: The track gauge W is obtained by calculating the distance between the right edge point of the left track and the left edge point of the right track in the yz plane. i ; The height of each left and right track: The distance from the median coordinate point of each track to the ground reference fitting line in the yz plane is obtained to obtain the track α height h i，α , rail β height h i,β ; Roll angle: The angle between the midpoints of the left and right rails and the ground reference fitting straight line in the yz plane is obtained to obtain the roll angle γ i ; The smoothness of the cross-sectional end surface of the left and right rails: a window is defined with the midpoint of each rail as the center, and the zy fitting line of the coordinate points in the window is calculated. Then, the discreteness of these coordinate points relative to the fitting line is calculated to obtain the smoothness of the cross-sectional end surface of rail α. Rail β cross-section end surface smoothness Pitch angle of left and right rails: Calculate the arctangent of the height change of the median coordinate point and the x difference between the two frames before and after to obtain the pitch angle θ of rail α i，α , track β track pitch angle θ i，β ; Left and right track smoothness: Calculate the zx fitting line with the current frame and the previous frames as the regression point set, and then calculate the discreteness of these coordinate points relative to the fitting line to obtain the track α smoothness and track β track smoothness 6. The track anomaly detection device according to claim 1, characterized in that: In step (4), segmented detection is performed by marking to facilitate the determination of the location where the abnormality occurs, specifically: The identification reading device (202) fixed on the carrier reads the identification corresponding serial number, the identification is arranged at equal intervals of Δx on the outside of the track, and the abnormal frame is identified at the time t when identification s and identification s+1 are read. s and t s+1 Between, where s = I, II, III..., and from t s The calculation starts when the marker s is read at time t. The mth frame of linear array data has parameter anomalies, i.e., the anomaly recognition time t m Corresponding to the mth frame of linear point cloud data after the identification s is read, the abnormal position can be accurately located according to the speed information obtained by the speed sensor: Where v m represents the real-time speed of the vehicle (1) when acquiring the mth frame of data, t1 and v1 represent the time of the first frame from reading the tag s and the real-time speed of the vehicle (1), respectively, m represents the distance from the abnormal point to the same scanning surface as marker s, x m_s It represents the distance from the outlier point to the marker s, v i It indicates the real-time speed of the vehicle collected by the speed sensor when the i-th frame of linear point cloud data is obtained starting from the reading of the identifier s.

7. The track anomaly detection device according to any one of claims 1 or 6, characterized in that: In step (5), track anomalies and changes are detected by combining the A and B modes: Mode A: Threshold discrimination method, that is, if a key shape and position characteristic parameter of a track exceeds the normal threshold range, it can be considered that a track shape and position abnormality occurs at the corresponding position of the current scanning frame; Mode B: Benchmark template comparison method. Specifically, first complete a preliminary scan of all sections in the entire detection domain. According to the calculation method of the key shape and position feature parameters of the track, obtain the shape and position parameter vectors of each frame in each section to construct a benchmark template. Then, during the actual detection, use the interpolation method to calculate the difference vector between the shape and position parameter vector of the current detection frame and the benchmark template. Based on the difference vector, the size of the change in the corresponding features of each vector element compared to the initial benchmark can be determined, thereby judging the changes in the track shape and position on the time scale.

8. The track anomaly detection device according to claim 7, characterized in that: Benchmark template comparison method, specifically including: In any segment divided by the identifiers s and s+1, the 1st to nth frames of point cloud data are collected, and the shape and position parameter vectors S1, S2, ... S corresponding to each of these n frames of data are obtained. i , ... S n , forming a reference template, where any vector Among them, the first element x of the vector i From the x in formula 3 m_I The solution method is determined, represents the distance from the frame data to the marker s, and the remaining elements are all track shape and position parameters determined according to Table 1; In actual detection, when the vehicle passes through the segment divided by the identifiers s and s+1, the parameter vectors corresponding to all frame line array point cloud data in the segment are obtained, where a frame vector S m It is expressed as: If x i <x m <x i+1 , that is, S m The corresponding scanning position is between the reference template vector S i and S i+1 The difference vector ΔS is calculated by interpolation. m : ΔS m The difference between the actual detection and the initial benchmark is characterized, and the change in each track structure parameter is determined by its 2nd to 13th elements; For example, in actual detection, the difference vector ΔS corresponding to the mth frame of a certain scanning position m The fourth element ΔW m If the absolute value is greater than the specified value, it means that the x after the mark s at the corresponding scanning position m_s The track gauge at has a significant change compared to the initial reference. A positive value means that the track gauge has become larger, and a negative value means that the track gauge has become smaller. The seventh element Δγ m If a positive or negative change greater than the specified value occurs, it means that the roll angle at the corresponding position has changed significantly compared to the pre-scan. If the vehicle still passes through this place at the same speed, there is a risk of overturning, and the place needs to be recalculated. Therefore, mode B can accurately obtain the location, type and degree of the change, which is convenient for targeted maintenance and inspection.

Images