Road surface texture parameter acquisition method based on three-dimensional laser and encoder array

Abstract

The present invention discloses a method for collecting road surface texture parameters based on a three-dimensional laser and encoder array, including: S1. Elevation point cloud collection; S2. Abnormal data cleaning; S3. Vibration noise suppression; S4. Correction of non-parallel measurement lines; S5. Extraction of road surface three-dimensional texture parameters based on local peak top surfaces; S6. Construction of an automatic texture parameter collection system, encapsulation and merging of all programs in S1-5, obtaining an automatic collection system that inputs point cloud data and outputs various road surface three-dimensional texture parameters in a specified area. The present invention realizes the automation and intelligence of the entire process of laser point cloud technology from obtaining elevation point clouds to extracting road surface three-dimensional texture parameters, reducing labor costs and processing time.

Description

Technical Field

[0001] The present invention belongs to the field of detection technology, and in particular relates to a road surface texture parameter acquisition method based on a three-dimensional laser and encoder array. Background Art

[0002] The anti-skid performance of asphalt pavement has a direct impact on traffic safety, especially when the car is driving at high speed or accelerating or decelerating. The friction between the tire and the road depends on the texture structure formed by the asphalt, aggregate and the small undulations on the aggregate surface. Therefore, accurately obtaining the texture structure and quantifying it with parameters to indirectly characterize the anti-skid performance of the road is one of the main methods for pavement service performance management.

[0003] Pavement texture can be obtained by direct and indirect methods. Direct methods include imaging, CT scanning, mechanical stylus and laser ranging, while indirect methods include sanding, drainage and blade methods.

[0004] The image method is based on various image processing technologies, partially combined with machine learning technology, to infer and restore two-dimensional image data into three-dimensional spatial data. The image method has a simple equipment architecture, but the upper limit of accuracy is low, and it has high requirements for detection conditions such as lighting, which limits the detection efficiency. The CT scanning method has good accuracy, but the CT scanner must be used under radiation protection conditions and has high energy consumption, so this technology is only suitable for indoor research. The mechanical stylus method slides a probe across the road surface with a certain pressure, and extracts the macro texture by obtaining the displacement of the probe, with high accuracy. However, the measurement range of this method is limited, the measurement results are greatly affected by the properties of the probe itself, and the probe is easily damaged due to frequent friction with the road surface, so this method has not been widely used.

[0005] The sand spreading method obtains the paving area by evenly spreading a quantitative volume of sand on the road surface, and the ratio of the volume of sand to the paving area is used to represent the average structural depth of the measuring point. The electric sand spreading method reduces the subjective error caused by manual operation to a certain extent, but it does not significantly improve the efficiency. The drainage method indirectly reflects the macroscopic structure of the road surface by monitoring the time it takes for a fixed volume of water to be discharged through the road surface. Whether it is the manual sand spreading method, the electric sand spreading method or the drainage method, it requires a closed inspection section and can only be inspected point by point. The blade method obliquely projects the laser beam at a 45° angle on the asphalt concrete specimen, extracts the light and dark boundaries formed to obtain the two-dimensional texture contour, but this method has strict requirements on the lighting conditions of the environment and is not suitable for field testing.

[0006] As an advanced measurement method, 3D laser point cloud has the advantages of fast, real-time, high density, high precision, dynamic and active. Fixing the laser point cloud instrument and its supporting equipment on the car constitutes a mobile measurement system, which can quickly and high-resolution obtain the road elevation point cloud without affecting traffic. The measurement system has the characteristics of low cost, labor saving, high safety and little interference with the working environment. However, it still has defects, such as large data noise and unreasonable texture parameter extraction method. Summary of the invention

[0007] The purpose of the present invention is to solve the above problems and propose a fully automated extraction method from road elevation point cloud to output three-dimensional texture parameters.

[0008] In order to solve at least one of the above technical problems, according to one aspect of the present invention, a method for collecting road texture parameters based on a three-dimensional laser and encoder array is provided, comprising the following steps:

[0009] S1. Elevation point cloud collection;

[0010] S2. Abnormal data cleaning;

[0011] S3. Vibration noise suppression;

[0012] S4. Correction of non-parallel measurement lines;

[0013] S5. Extraction of road surface 3D texture parameters based on local peak surface;

[0014] S6. Construct an automatic texture parameter acquisition system, encapsulate and merge all the programs in S1-5, and obtain an automatic acquisition system that inputs point cloud data and outputs various road surface three-dimensional texture parameters in a specified area.

[0015] Furthermore, in S1, a three-dimensional laser scanner and a traveling encoder array are integrated on the vehicle;

[0016] A fixed bracket is installed at the rear of the vehicle, and a 3D laser scanner is set up. The 3D laser scanner is perpendicular to the road surface. The distance between the 3D laser scanner and the road surface is determined according to the scanning range of the 3D laser scanner, ensuring that the scanning range covers at least one road lane;

[0017] The traveling encoder array consists of two traveling encoders installed on the left rear and right rear wheels respectively; the encoder on the right rear wheel is used to accurately monitor the turning motion parameters of the vehicle, which is called the control encoder; the encoder on the left rear wheel is used to accurately measure the vehicle's travel distance and control the 3D laser scanner data storage, which is called the control encoder. Every time the vehicle travels W meters, the control encoder triggers the saving function of the 3D laser scanner to save the 3D scanning results of W meters of road surface in the form of a point cloud data file in .dat format.

[0018] Furthermore, in S2, the abnormal data cleaning is specifically as follows: taking the dat file as input, decoding the dat file into a matrix of size W×H, W is the vehicle travel length covered by the point cloud data file, H is the scanning width covered by the point cloud data file, and saving the height Z measured by the laser point cloud within the W×H range in the corresponding position; the automatic conversion program is based on the normal distribution fitting method, and the mean and variance of the elevation in a certain matrix are statistically calculated, and the elevation points outside the confidence interval are considered to be abnormal points, and the elevation of the abnormal point is assigned to the average value of the elevations of the points around the abnormal point.

[0019] Furthermore, in S3, the vibration noise suppression is specifically:

[0020] Set up a vibration noise suppression program based on the "reference surface assumption", including three sub-modules: determining the reference surface, determining the reference line, and regressing the reference line;

[0021] Among them, the "reference plane hypothesis" believes that if the plane with the smallest distance to all points is defined as the reference plane, then in a point cloud matrix with a sufficient number of measurement lines, the reference plane obtained based on this point cloud matrix is ​​the same before and after the vertical vibration occurs;

[0022] Among them, the reference surface determination module is completed based on the numerical solution;

[0023] The module for determining the baseline is completed based on the numerical solution; the baseline is the straight line on a certain survey line with the smallest sum of distances to all points;

[0024] The baseline regression module synchronously regresses the baseline and the measuring point to the baseline plane by keeping the position of each point relative to the baseline unchanged.

[0025] Furthermore, in S4, the non-parallel correction of the survey line is specifically as follows:

[0026] Set the alignment correction procedure based on the encoder array, the number of encoder triggers , trigger frequency , Advance Distance ;

[0027] Take a short period of time during the turning process, in which the motion trajectory of the two rear wheels is considered to be a circular arc, and assume that the triggering times of the outer encoder and the inner encoder in the short period of time are , The trigger frequency of both is ;

[0028] , are the distances traveled by the outer wheel and the inner wheel in the same time, , are the turning radius of the outer wheel and the inner wheel respectively, , Known, and The difference is the wheelbase, which is also known, and can be solved , According to the relationship between arc length, radius and radian, the angle of rotation of the two rear wheels in the short period of time is obtained. , assuming that the control encoder is located on the inside and the reference encoder is located on the outside, then the innermost point of the measuring line is Spacing in direction It is equal to the distance when the detection vehicle is driving in a straight line. The outermost point of the measurement line is The spacing in the direction is ; During the short period of time, the remaining points on the survey line are The spacing in the direction is ; is the number of measuring points on a measuring line; the turning process is divided into several small time periods, and the deflection of the measuring line in each small time period is analyzed respectively, so as to obtain the real position of the road surface point cloud collected during the turn in the physical space.

[0029] Furthermore, in S5, the specific steps of extracting the three-dimensional texture parameters of the road surface based on the local peak surface are:

[0030] The peak surface is defined as the plane of the highest three points in the measured point cloud matrix. First, the point cloud matrix to be tested is divided into several small matrices, and then the peak surface of each small matrix is ​​calculated respectively. The texture parameters in the measured area are calculated based on this peak surface, and finally the average value of the texture parameters obtained from all small matrices is calculated.

[0031] According to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the steps of the road surface texture parameter acquisition method based on a three-dimensional laser and encoder array of the present invention are implemented.

[0032] According to another aspect of the present invention, there is provided a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the road surface texture parameter acquisition method based on a three-dimensional laser and encoder array of the present invention when executing the program.

[0033] Compared with the prior art, the above method of the present invention has the following beneficial effects:

[0034] The present invention realizes the automation and intelligence of the whole process of laser point cloud technology from obtaining elevation point cloud to extracting three-dimensional texture parameters of road surface, thereby reducing labor costs and processing time. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0036] Figure 1 This is a schematic diagram of the reference plane solution result of an embodiment of the present invention;

[0037] Figure 2 This is a schematic diagram of the vibration noise suppression result of an embodiment of the present invention;

[0038] Figure 3 It is a schematic diagram of a local peak top surface of an embodiment of the present invention. DETAILED DESCRIPTION

[0039] To make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the embodiments of the present invention will be described clearly and completely in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them.

[0040] Unless otherwise defined, technical or scientific terms used herein shall have the same meaning as commonly understood by one of ordinary skill in the art to which the invention belongs.

[0041] like Figure 1-Figure 3 As shown in the figure, a method for collecting 3D texture parameters of road surface based on 3D laser point cloud and encoder array realizes the automation and intelligence of the whole process of laser point cloud technology from obtaining elevation point cloud to extracting 3D texture parameters of road surface, reduces labor costs and processing time. It specifically includes the following steps:

[0042] S1. Elevation point cloud acquisition: 3D laser scanner and travel encoder array are integrated on the vehicle.

[0043] A fixed bracket is installed at the rear of the vehicle, and a three-dimensional laser scanner is set up. The three-dimensional laser scanner is perpendicular to the road surface. The distance between the three-dimensional laser scanner and the road surface is determined according to the scanning range of the three-dimensional laser scanner to ensure that the scanning range covers at least one road lane.

[0044] The traveling encoder array consists of two traveling encoders installed on the left rear and right rear wheels respectively; the encoder on the right rear wheel is used to accurately monitor the turning motion parameters of the vehicle, which is called the control encoder; the encoder on the left rear wheel is used to accurately measure the vehicle's travel distance and control the 3D laser scanner data storage, which is called the control encoder. Every time the vehicle travels W meters, the control encoder triggers the saving function of the 3D laser scanner to save the 3D scanning results of W meters of road surface in the form of a point cloud data file in .dat format;

[0045] S2. Abnormal data cleaning: The present invention has developed a set of abnormal data cleaning programs, which takes the dat file as input, and decodes it into a matrix of size, W is the vehicle travel length covered by the point cloud data file, H is the scan width covered by the point cloud data file, and saves the height Z measured by the laser point cloud within the range of W×H in the corresponding position; the automatic conversion program is based on the normal distribution fitting method to count the mean of the elevation in a certain matrix With variance , which is considered to be within the confidence interval The elevation points outside the point are outliers, and the elevation of the outlier point is assigned the average elevation of the points around it, that is:

[0046]

[0047] In the formula, After data cleaning, is the original data.

[0048] S3. Vibration noise suppression: The present invention develops a vibration noise suppression program based on the "reference plane assumption", including three sub-modules: determining the reference plane, determining the reference line, and baseline regression.

[0049] Among them, the "reference plane hypothesis" believes that if the plane with the smallest distance to all points is defined as the reference plane, then in a point cloud matrix with a sufficient number of measuring lines, the reference plane obtained based on this point cloud matrix is ​​the same before and after the vertical vibration occurs.

[0050] Among them, the reference surface determination module is completed based on numerical solution.

[0051] Assume the reference plane equation is ,make

[0052] Then the equation for the sum of the squares of the distances from all points to the line can be simplified to:

[0053]

[0054] Therefore, the problem is transformed into a Lagrange extreme value problem, that is, finding the function The extreme point of:

[0055]

[0056] in is the point cloud coordinate. Find the partial derivatives and set all partial derivatives equal to 0:

[0057]

[0058] From (7), we get:

[0059]

[0060] Substituting (9) into (4)-(6) yields:

[0061]

[0062] Arrange (10)-(12) into matrix form:

[0063]

[0064] Assume that the square matrix in the above formula is matrix A, and the vector is , then the above formula can be rewritten as:

[0065]

[0066] Right now

[0067]

[0068] Formula (15) is exactly the same as the definition of matrix eigenvalue-eigenvector, so the three elements in the eigenvector corresponding to the minimum eigenvalue of matrix A are , , On this basis, calculate the distance between each point and the reference plane and find all distances Standard Deviation , remove points, and recalculate using the remaining points , , . Repeat this elimination step until all remaining points are within .

[0069] The module for determining the baseline is completed based on the numerical solution. The baseline is the straight line with the minimum distance to all points on a certain survey line. The baseline is determined based on the numerical solution, and the final expression is:

[0070]

[0071] The baseline regression module regresses the baseline and the measured point to the baseline plane synchronously by keeping the position of each point relative to the baseline unchanged, that is:

[0072]

[0073] Where: is the point cloud data after noise suppression, To be the first The point cloud is located , The solution after substituting the coordinates into the reference plane equation is: To be the first The point cloud is located , The solution result after substituting the coordinates into the corresponding baseline equation.

[0074] S4. Line non-parallel correction: The present invention develops a line non-parallel correction program based on an encoder array. The number of encoder triggers , trigger frequency , Advance Distance The relationship between can be expressed as:

[0075]

[0076] Take a short period of time during the turning process, during which the motion trajectory of the two rear wheels can be considered to be a circular arc. Suppose the triggering times of the outer encoder and the inner encoder in this short period of time are , The trigger frequency of both is According to formula (18) and the relationship between radius, arc length and radian, we can get:

[0077]

[0078] In the formula, and are the distances traveled by the outer wheel and the inner wheel in the same time, and are the turning radius of the outer wheel and the inner wheel respectively, , Known, and The difference is the wheelbase, which is also known, so we can solve , According to the relationship between arc length, radius and radian, we can get:

[0079]

[0080] In the formula is the rotation angle of the two rear wheels in this short period of time. Assuming that the control encoder is located on the inside and the reference encoder is located on the outside, the innermost point of the measuring line is The spacing in the direction is equal to the spacing when the detection vehicle is driving in a straight line The outermost point of the measuring line is Spacing in direction It can be expressed as:

[0081]

[0082] Furthermore, during this short period of time, the remaining points on the survey line Spacing in direction It can be expressed as:

[0083]

[0084] The formula is the number of measuring points on a measuring line. By dividing the turning process into several small time periods, and analyzing the deflection of the measuring line respectively, the real position of the road surface point cloud collected during the turn in the physical space can be obtained.

[0085] S5. Extraction of three-dimensional texture parameters of road surface based on local peak top surface: The peak top surface is defined as the plane passing through the three highest points in the point cloud matrix. The present invention imitates the test logic of the sand paving method. First, the point cloud matrix to be tested is divided into several small matrices. Then, the peak top surface of each small matrix is ​​calculated respectively. The texture parameters in the area are calculated based on this peak top surface, and finally the average value of the texture parameters obtained from all small matrices is calculated.

[0086] S6. Constructing an automatic texture parameter acquisition system: Encapsulate and merge all the aforementioned programs to obtain an automatic acquisition system that inputs point cloud data and outputs various road surface three-dimensional texture parameters in a specified area.

[0087] Embodiment: (1) Elevation point cloud acquisition: A three-dimensional laser scanner and a traveling encoder array are integrated on a vehicle; a clear day is selected to calibrate the three-dimensional laser scanner; the inspection vehicle is driven for 3 km on Jinyin Avenue in Nanjing, and 680 valid dat point cloud data files are obtained.

[0088] (2) Abnormal data cleaning: Taking the dat file in (1) as input, the program decodes it into a matrix of size 3639×2665×. The vehicle travel length covered by a single point cloud data file is 3639 pixels (i.e., 3639 survey lines), and the scanning width covered by a single point cloud data file is 2665 pixels (i.e., there are 2665 survey points on a survey line). It is a certain elevation measured by the laser point cloud within the range of 3639×2665; the mean and standard deviation of the elevations of all points in the matrix are 0.127. Therefore, the elevations of all points with elevations lower than 1.646 or higher than 2.408 need to be replaced with the average of the elevations of their surrounding points.

[0089] (3) Vibration noise suppression: Solve the reference plane of the matrix. The solution is as follows: Figure 1 As shown, the baseline of each measurement line is solved separately, and the vibration noise is suppressed by regression. The suppression result is shown in Figure 2 shown.

[0090] (4) Correction of non-parallel survey lines: Based on the data sent back by the control encoder and the reference encoder in the point cloud file segment, it is calculated that the survey line 1464-2397 is a turning section. The rotation angle between each two survey lines in the turning section is calculated, and the distance between the survey lines is corrected.

[0091] (5) Extraction of road surface 3D texture parameters based on local peak top surface: Divide the matrix into blocks and calculate the peak top surface separately. Figure 3 , taking the mean texture depth (MTD) as an example, the MTD of each block is calculated, which are 0.9446, 0.9509, 0.8432, 0.9443, 0.8276, 0.7871, 1.0124, 0.7449, 0.9895,0.9826, 0.9679, 0.8669, 0.7551, 0.8741, 0.8833, 0.7891, 0.9835, 0.7738,0.9041, 0.7221, 0.9003, 0.7233, 0.9232, 0.9839, 0.8381. The average value is 0.8767.

[0092] (7) Constructing an automatic texture parameter acquisition system: Encapsulate and merge all the aforementioned programs to obtain an automatic acquisition system that inputs point cloud data and outputs various road surface three-dimensional texture parameters in a specified area.

[0093] The examples described in the present invention are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the concept and scope of the present invention. Without departing from the design concept of the present invention, various modifications and improvements made to the technical solutions of the present invention by engineers and technicians in this field should all fall within the protection scope of the present invention.

Claims

1. A road surface texture parameter acquisition method based on a three-dimensional laser and encoder array, characterized in that: The steps include: S1. Elevation point cloud collection; S2. Abnormal data cleaning; S3. Vibration noise suppression; S4. Correction of non-parallel measurement lines; S5. Extraction of road surface 3D texture parameters based on local peak surface; S6. Construct an automatic texture parameter acquisition system, encapsulate and merge all the programs in S1-5, obtain an automatic acquisition system that inputs point cloud data and outputs various road surface three-dimensional texture parameters in a specified area; In S3, vibration noise suppression is specifically as follows: Set up a vibration noise suppression program based on the "reference surface assumption", including three sub-modules: determining the reference surface, determining the reference line, and regressing the reference line; Among them, the "reference plane hypothesis" believes that if the plane with the smallest distance to all points is defined as the reference plane, then in a point cloud matrix with a sufficient number of measurement lines, the reference plane obtained based on this point cloud matrix is ​​the same before and after the vertical vibration occurs; Among them, the reference surface determination module is completed based on the numerical solution; The module for determining the baseline is completed based on the numerical solution; the baseline is the straight line on a certain survey line with the smallest sum of distances to all points; Among them, the baseline regression module synchronously regresses the baseline and the measured point to the baseline plane by keeping the position of each point relative to the baseline unchanged; In S4, the correction of non-parallel measurement lines is as follows: Set up the non-parallel measurement line correction program based on the encoder array, the encoder trigger times N, trigger frequency f, and advance distance S; Take a short period of time during the turning process, in which the motion trajectory of the two rear wheels is considered to be a circular arc, and assume that the number of triggering times of the outer encoder and the inner encoder in the short period of time are N1 and N2 respectively, and the triggering frequency of both is f; S1 and S2 are the distances traveled by the outer wheel and the inner wheel in the same period of time, r1 and r2 are the turning radii of the outer wheel and the inner wheel, respectively. N1 and N2 are known, and the difference between r1 and r2 is the wheelbase, which is also known. The values ​​of r1 and r2 can be solved. According to the relationship between arc length, radius and radian, the angle α of the rotation of the two rear wheels in the short period of time is obtained. Assuming that the control encoder is on the inside and the control encoder is on the outside, the spacing of the innermost points of the measuring line in the y direction is equal to the spacing Δy when the detection vehicle is driving in a straight line, and the spacing of the outermost points of the measuring line in the y direction is Δy. n ; During the short period of time, the spacing between the remaining points on the survey line in the y direction is Δy i ; n is the number of measuring points on a measuring line; the turning process is divided into several small time periods, and the deflection of the measuring line in each small time period is analyzed respectively, so as to obtain the real position of the road surface point cloud collected during the turn in the physical space.

2. The method according to claim 1, characterized in that In S1, a 3D laser scanner and a traveling encoder array are integrated on the vehicle; A fixed bracket is installed at the rear of the vehicle, and a 3D laser scanner is set up. The 3D laser scanner is perpendicular to the road surface. The distance between the 3D laser scanner and the road surface is determined according to the scanning range of the 3D laser scanner, ensuring that the scanning range covers at least one road lane; The traveling encoder array consists of two traveling encoders installed on the left rear and right rear wheels respectively; the encoder on the right rear wheel is used to accurately monitor the turning motion parameters of the vehicle, which is called the control encoder; the encoder on the left rear wheel is used to accurately measure the vehicle's travel distance and control the 3D laser scanner data storage, which is called the control encoder. Every time the vehicle travels W meters, the control encoder triggers the saving function of the 3D laser scanner to save the 3D scanning results of W meters of road surface in the form of a point cloud data file in .dat format.

3. The method according to claim 1, characterized in that In S2, the abnormal data cleaning is specifically as follows: taking the dat file as input, decoding the dat file into a matrix of size W×H, W is the vehicle travel length covered by the point cloud data file, H is the scanning width covered by the point cloud data file, and saving the height Z measured by the laser point cloud within the W×H range in the corresponding position; the automatic conversion program is based on the normal distribution fitting method, and the mean and variance of the elevation in a certain matrix are statistically calculated. The elevation points outside the confidence interval are considered to be abnormal points, and the elevation of the abnormal points is assigned to the average value of the elevations of the points around the abnormal points.

4. The method according to claim 1, characterized in that In S5, the specific steps of extracting the three-dimensional texture parameters of the road surface based on the local peak surface are: The peak surface is defined as the plane of the highest three points in the measured point cloud matrix. First, the point cloud matrix to be tested is divided into several small matrices, and then the peak surface of each small matrix is ​​calculated respectively. The texture parameters in the measured area are calculated based on this peak surface, and finally the average value of the texture parameters obtained from all small matrices is calculated.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the road surface texture parameter acquisition method based on a three-dimensional laser and encoder array as described in any one of claims 1 to 4 are implemented.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the road surface texture parameter acquisition method based on a three-dimensional laser and encoder array as described in any one of claims 1 to 4 are implemented.

Images