A road geometry information extraction method based on laser point cloud

Abstract

The present invention discloses a method for extracting road geometry information based on laser point cloud, including: performing radius filtering and grid downsampling on point cloud to simplify point cloud; taking into account that the elevation distribution of road surface points is relatively concentrated and the road surface is smoother, extracting elevation features and local normal vector features of point cloud to distinguish ground points from non-ground points; using regional growing method to connect road surface points on ground points, and obtaining a complete road surface point cloud after restoring the mistakenly deleted points; finally, according to the collected vehicle trajectory information, calculating the trajectory vector and cutting the road cross section, and using the least squares method to obtain the geometric parameters of the road. The present invention takes into account both extraction accuracy and program operation efficiency, and can comprehensively consider the automatic extraction of road geometry information under different road environments.

Description

Technical Field

[0001] The invention relates to the technical field of automatic road surface detection, and in particular to a road geometry information extraction algorithm based on laser point cloud. Background Art

[0002] Since the reform and opening up, my country's road network has expanded rapidly. The amount of road reconstruction and expansion projects has increased year by year, and over time, the roadbed is prone to uneven settlement. These reasons have made the actual situation of the road more and more complicated. How to collect road information efficiently and at low cost has become an important issue in road maintenance and management. In recent years, the fields of autonomous driving, smart transportation, and high-precision map production have flourished, and it is more urgent to accurately and efficiently grasp road information. The workload of manually collecting road characteristic parameters is large, and it is difficult to carry out on a large scale, and it requires a lot of manpower, financial resources, and material resources, which has prompted the extraction of road information to develop in the direction of automation.

[0003] The vehicle-mounted laser scanning technology installs a laser scanner on the car. When the car moves, it can collect the surrounding three-dimensional surface information along the driving path, and generate a three-dimensional laser point cloud according to the coordinates and reflection intensity. The vehicle-mounted laser scanning system has ushered in rapid development and maturity in recent years. As a rapidly developing mobile measurement technology, compared with traditional manual measurement technology, it can quickly, accurately and widely obtain the intensity information, three-dimensional spatial information and even color information of roads and road environments, which can play an important role in the later management and maintenance of roads and smart cities. Due to the huge amount of point cloud data, improving the computational efficiency of the algorithm is very important for the large-scale application of vehicle-mounted laser scanning systems. In addition, the road environment is complex and changeable, and there may be complex road scenes such as intersections and tunnels. Regarding the extraction of laser radar point cloud data, there is still a lack of professional traffic processing of the extraction results. The horizontal, vertical and horizontal parameters that are very important in the field of road design and maintenance management have not yet been involved. Summary of the invention

[0004] Purpose of the invention: In view of the current situation that manual collection of road geometry information is time-consuming and labor-intensive, a road geometry information extraction algorithm based on laser point cloud is provided, which can comprehensively consider the automated extraction of road geometry information under different road environments and optimize the existing road geometry information extraction algorithm.

[0005] Technical solution: To achieve the above-mentioned invention object, the present invention adopts the following technical solution:

[0006] A road geometry information extraction algorithm based on laser point cloud, comprising the following steps:

[0007] (1) Perform radius filtering and grid downsampling on the point cloud collected by the vehicle-mounted laser scanning system to simplify the point cloud;

[0008] (2) Extract the elevation features and local normal vector features of the simplified point cloud to distinguish ground points from non-ground points;

[0009] (3) Based on the three-dimensional coordinates of the point cloud and the laser reflection intensity information, the region growing method is used on the ground points to connect the road surface points and remove the non-road surface parts of the ground points. After restoring the mistakenly deleted points, a complete road surface point cloud is obtained;

[0010] (4) Cutting a road cross section from the road surface point cloud obtained in (3) according to the vehicle trajectory information;

[0011] (5) The road width and longitudinal slope are calculated using the cross section, and the road curvature and transverse slope are calculated using the least squares method to complete the extraction of road geometry information.

[0012] Furthermore, in step (1), the method of performing radius filtering and grid downsampling on the point cloud to simplify the point cloud is as follows: first, the KD Tree method is used to establish a data point index; second, for each point, all points within a distance R are searched, and if the total number is less than a threshold K, they are deleted, and the search distance R = 1.0m, and the radius filter threshold K = 100; then the point cloud is divided using a fixed-size three-dimensional grid, each grid is a cube with an edge length of m, and the centroid of the grid is selected to replace all the points of the grid, and the value of the edge length m is 0.4;

[0013] Furthermore, the method for extracting the elevation features and local normal vector features of the point cloud in step (2) to distinguish ground points from non-ground points is:

[0014] (a) Create a two-dimensional grid on the horizontal plane, project the simplified point cloud onto the horizontal plane and divide the point cloud into many squares according to the grid, restore the elevation information of the point cloud in each square, and obtain many continuously distributed three-dimensional vertical columns. After that, the operation is performed on each vertical column one by one;

[0015] (b) Rearrange the point cloud in the vertical column in descending order of elevation value Z, and take the average elevation value of the lowest 100 points as the lowest elevation Z. min , only keep the elevation value in Z min arrive Points within the range, where is the maximum allowable variation range of ground point elevation, set to 1.0m;

[0016] (c) For each point in the data set, find the 20 points closest to it to construct a local point set and perform standardization. The formula is:

[0017]

[0018] where xi ,y i and z i are the X, Y, and Z coordinate values ​​of the i-th point in the local point set, Respectively represent the average values ​​of the X, Y, and Z coordinates in the point set, and Var(x), Var(y), and Var(z) represent the variances of the X, Y, and Z coordinates in the point set. Calculate the correlation coefficient matrix, which is:

[0019]

[0020] in, x, y and z represent the X, Y and Z coordinate vectors of the local point set respectively. Cov represents the covariance. The eigenvector corresponding to the minimum eigenvalue is the normal vector of the point cloud. Points whose angles between the normal vector and the Z axis are greater than 5° are deleted.

[0021] Furthermore, in step (3), the specific method of using the region growing method on the ground points to connect the road points and remove the non-road parts of the ground points is as follows: first, a point is randomly selected from the ground points as a seed point as the starting point for the growth of this region, and then the neighboring points of this seed point are compared to see whether they meet the growth criterion, which is:

[0022] (a) The distance from the seed point is less than the distance threshold D, assuming D = 1.0m;

[0023] (b) The elevation difference of the seed point is less than the elevation difference threshold, and the elevation difference threshold is set to 0.03m;

[0024] (c) The reflection intensity difference with the seed point is less than the intensity difference threshold, and the intensity difference threshold is set to 50;

[0025] The neighboring points that meet the growth criteria are included in the region and used as new seed points of this region to continue growing outward until all points with similar spatial distribution characteristics and laser reflection intensity characteristics are included in this region; if the number of points in this region exceeds 50% of the total number of ground points, then this region is the road surface part of the ground points, otherwise this region is the non-road surface part of the ground points; select non-road surface points one by one and compare their Euclidean distances (d Δ ) and elevation difference (h Δ ), if the Euclidean distance and elevation difference between the non-road point and a road point are within the range of D / 2 and 0.03m, then the non-road point is restored to a road point;

[0026] Furthermore, in step (4), based on the collected vehicle trajectory information, the vehicle's trajectory points are used to define the position vectors of the starting point and the end point of each trajectory vector, and the trajectory vector with the i trajectory point as the starting point is It can be expressed as: in and is the position vector of the starting and ending points of the trajectory vector; then calculate each trajectory vector The normal vector The expression is:

[0027]

[0028] Among them, v x 、v y 、v z is the trajectory vector The three elements represent the projection length of the trajectory on the X, Y, and Z axes respectively; then a cuboid with a height of 5.0m and a thickness of 1.5m is used to cut the road cross section along the normal vector direction of the trajectory point; finally, the rotation matrix is ​​used to rotate the cross section to the positive direction of the Y axis. The rotation matrix is ​​as follows:

[0029]

[0030] Where θ is the trajectory vector The rotation angle in the clockwise direction of the positive Y axis.

[0031] Furthermore, in step (5), the specific method of calculating the road width and longitudinal slope by using the cross section and the least square method to calculate the road curvature and transverse slope is as follows:

[0032] (a) In each cross section, the road width at that cross section is obtained by subtracting the minimum abscissa x_min from the maximum abscissa x_max;

[0033] (b) Calculate the cross section CS corresponding to trajectory point i i The horizontal coordinate mean x_mean, the vertical coordinate mean y_mean and the elevation mean z_mean, then the point with (x_mean, y_mean, z_mean) as the coordinate That is the cross section CS i The road center point at the location; the road center point set obtained by calculating all trajectories constitutes the road center line, and the road center line elevation is used to calculate the road longitudinal slope l, which is expressed as: Among them, H is the relative height difference between the two points before and after the road centerline, and L is the horizontal distance between the two points before and after the road centerline;

[0034] (c) For points on the road centerline Let A be the search range when calculating the curvature, find all the road center points within the radius A / 2, let A = 10m, these points form an arc, and the corresponding circle equation is: 2 +y 2+ax+by+c=0, where x, y, z are the coordinates of the center point of the road, and a, b, c are the characteristic parameters of the circle. Use the least squares method to fit the circle and find a, b, c that minimizes the sum of squares of the residuals. Then you can calculate the radius r of the standard circle, and its reciprocal 1 / r is the curvature.

[0035] (d) Taking the center line of the road as the dividing point, the road surface is divided into two point sets on the left and right, and the least squares method is used to fit the straight line to obtain the slope, which is the cross slope.

[0036] Beneficial effects: Compared with the prior art, the present invention solves the shortcomings of the existing road geometry information acquisition methods that are time-consuming and labor-intensive on the basis of the prior art. At the same time, the method disclosed by the present invention is simple and convenient, and the results obtained fill the gap in the prior art in this field, thereby being able to comprehensively consider the automated extraction of road geometry information under different road environments and optimize the existing road geometry information extraction algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flow chart of the method provided by the present invention;

[0038] Figure 2 Schematic diagram of the region growing method;

[0039] Figure 3 Cut schematic diagram for road cross section; DETAILED DESCRIPTION

[0040] The present invention will be further explained below in conjunction with the accompanying drawings.

[0041] like Figure 1 As shown, the present invention is a road geometry information extraction algorithm based on laser point cloud, which mainly includes the following steps:

[0042] (1) Perform radius filtering and grid downsampling on the point cloud collected by the vehicle-mounted laser scanning system to simplify the point cloud.

[0043] Mobile laser scanning technology is to install a laser scanner on a car. When the car is moving, it can quickly obtain the three-dimensional surface information around the driving path and generate a three-dimensional laser point cloud based on the X, Y, Z coordinates and the intensity value I. The laser beam of the mobile laser scanning system is dense and flexible, and the pulse emission frequency can reach hundreds of thousands of points per second. The three-dimensional laser point cloud established has a large amount of data. The ultra-high point density makes the point cloud file larger, and the processing and rendering of the point cloud file are relatively slow. In addition, the point cloud is usually composed of disorganized data points, and there is no continuous position relationship between the data points. The order of the data points reflects the order of scanning, but cannot represent the spatial position information between the points.

[0044] This method uses KD Tree to establish a point cloud index. KD Tree is a tree data structure that stores instance points in k-dimensional space for fast retrieval. It is mainly used for searching key data in multidimensional space. Each node of K-DTree is a binary tree of k-dimensional points. All non-leaf nodes can be regarded as a hyperplane that divides the space into two half-spaces. The subtree to the left of the node represents the points on the left side of the hyperplane, and the subtree to the right of the node represents the points on the right side of the hyperplane. The method for selecting a hyperplane is as follows: each node is related to the dimension in k dimensions that is perpendicular to the hyperplane. Therefore, if you choose to divide according to the x-axis, all nodes with x values ​​less than the specified value will appear in the left subtree, and all nodes with x values ​​greater than the specified value will appear in the right subtree. In this way, the hyperplane can be determined by the x value, and its normal is the unit vector of the x-axis.

[0045] In order to simplify the point cloud, this method uses radius filtering and grid downsampling. The radius filter creates a sphere with a certain radius centered on a certain point, and then calculates the number of points in the data set that are inside this sphere. When the number of points meets the threshold condition, the point is retained, and when it does not meet the threshold condition, the point is deleted. Subsequently, the original point cloud is divided into a spatial grid, and the center of gravity of all points in each cube grid is calculated. The center of gravity point is used to replace all points in this cube grid, that is, the coordinate average, intensity average, and color information average of all points in the grid are merged and simplified into one point. The grid downsampling method can better preserve the geometric features of the original point cloud, significantly reduce the point cloud density, and improve computational efficiency.

[0046] (2) Extract the elevation features and local normal vector features of the simplified point cloud to distinguish ground points from non-ground points.

[0047] Compared with objects such as cars and trees, the elevation distribution of the road surface is more concentrated and close to the ground. The most obvious geometric feature of the road surface is that it has a flat and continuous surface. The point cloud elevation feature and the local normal vector feature can be used to roughly extract the road surface.

[0048] (a) Create a two-dimensional grid on the horizontal plane, project the point cloud onto the horizontal plane and divide the point cloud into a number of squares according to the grid, and obtain a number of continuously distributed three-dimensional vertical columns. After that, the operation is performed on each vertical column one by one;

[0049] (b) Rearrange the point cloud in the vertical column in descending order of elevation value Z, and take the average elevation value of the lowest 100 points as the lowest elevation Z. min , only keep the elevation value in Z min arrive Points within the range, where is the maximum allowable range of ground point elevation, 1.0m;

[0050] (c) For each point in the data set, find the 20 points closest to it to construct a local point set and perform standardization. The formula is:

[0051]

[0052] where x i ,y i and z i are the X, Y, and Z coordinate values ​​of the i-th point in the local point set, Respectively represent the average values ​​of the X, Y, and Z coordinates in the point set, and Var(x), Var(y), and Var(z) represent the variances of the X, Y, and Z coordinates in the point set. Calculate the correlation coefficient matrix, which is:

[0053]

[0054] in, x, y and z represent the X, Y and Z coordinate vectors of the local point set respectively. Cov represents the covariance, and the eigenvalues ​​α1, α2 and α3 of the correlation coefficient matrix are further calculated, where α1>α2>α3. Then, the eigenvectors corresponding to α1, α2 and α3 are calculated. Normally, the eigenvalues ​​of road surface points are 0.4≤α1≤0.6, 0.4≤α2≤0.6, and 0≤α3≤0.1, and the plane features are more obvious. The eigenvector corresponding to the third principal component is the normal vector of the point cloud. Points whose angles between the normal vector and the Z axis are greater than 5° are deleted;

[0055] (3) The region growing method is used on the ground points to connect the road surface points, and the complete road surface point cloud is obtained after restoring the accidentally deleted points.

[0056] like Figure 2 As shown in the figure, the core idea of ​​the region growing method is to compare the similarity of the point cloud in the region with its neighboring points in terms of properties, so as to merge point clouds with similar properties. The specific method of using the region growing method on the ground points to connect the road points and remove the non-road parts of the ground points is as follows: first, a point is randomly selected from the ground points as a seed point as the starting point for the growth of this region, and then the neighboring points of this seed point are compared to see whether they meet the growth criteria. The growth criteria are:

[0057] (a) The distance from the seed point is less than the distance threshold D, assuming D = 1.0m;

[0058] (b) The elevation difference of the seed point is less than the elevation difference threshold, and the elevation difference threshold is set to 0.03m;

[0059] (c) The reflection intensity difference with the seed point is less than the intensity difference threshold, and the intensity difference threshold is set to 50;

[0060] The neighboring points that meet the growth criteria are included in the region and used as new seed points in this region to continue growing outward until all points with similar spatial distribution characteristics and laser reflection intensity characteristics are included in this region; if the number of points in this region exceeds 50% of the total number of ground points, then this region is the road surface part of the ground points, otherwise this region is the non-road surface part of the ground points; due to the gridding operation, there are fewer road surface points in the grid at the edge of the road, and some road surface points in such grids are not successfully extracted, so the extracted road edge is relatively rough, and the mistakenly deleted points need to be restored. Select non-road surface points one by one and compare their Euclidean distances (d Δ ) and elevation difference (h Δ ), if the Euclidean distance and elevation difference between the non-road point and a road point are within the range of D / 2 and 0.03m, then the non-road point is restored to a road point;

[0061] (4) Based on the collected vehicle trajectory information, a road cross section is cut from the road surface point cloud obtained in step (3).

[0062] The road is a strip-shaped entity laid flat in three-dimensional space. The line formed by the center of the road width is the road centerline. The road centerline is the core element of the road's horizontal, vertical and horizontal information. If you take any point on the centerline, the normal section at that point is the road cross section, and the vertical section of the road along the centerline is the longitudinal section of the route. The vehicle's trajectory points are used to define the position vectors of the starting and ending points of each trajectory vector. The trajectory vector with the i trajectory point as the starting point is It can be expressed as: in and are the position vectors of the starting and ending points of the trajectory vector, where vector Can be written as x i ,y i 、z i is the projection length of the starting position vector on the X, Y, and Z axes; i+1 ,y i+1 、z i+1 is the projection length of the end position vector on the X, Y, and Z axes. The number of trajectory vectors depends on the number of trajectory points of the vehicle when the lidar information is collected. and is the position vector of the starting and ending points of the trajectory vector; then calculate each trajectory vector The normal vector The expression is:

[0063]

[0064] Among them, v x 、v y 、v z is the trajectory vector The three elements represent the projection length of the trajectory on the X, Y, and Z axes respectively; then a cuboid with a height of 5.0m and a thickness of 1.5m is used to cut the road cross section along the normal vector direction of the trajectory point, such as Figure 3 As shown; finally, the cross section is rotated to the positive direction of the Y axis using the rotation matrix. The rotation matrix is ​​as follows:

[0065]

[0066] Where θ is the trajectory vector The rotation angle in the clockwise direction of the positive Y axis.

[0067] (5) Use cross sections to calculate road width and longitudinal slope, and use the least squares method to calculate road curvature and transverse slope.

[0068] (a) In each cross section, the road width at that cross section is obtained by subtracting the minimum abscissa x_min from the maximum abscissa x_max;

[0069] (b) Generally speaking, the design elevation of high-grade highways is the elevation at the outer edge of the central dividing strip, and the design elevation of rebuilt highways is the elevation of the road centerline or the elevation of the outer edge of the central dividing strip. In order to simplify the procedure, this method uses the elevation of the road centerline to calculate the longitudinal slope of the road. Calculate the mean value of the horizontal coordinate x_mean, the mean value of the vertical coordinate y_mean, and the mean value of the elevation z_mean of the cross section, then the coordinate point (x_mean, y_mean, z_mean) can represent the road center point at the slice. The road center point set calculated from all trajectories constitutes the road center line. The road center line elevation is used to calculate the road longitudinal slope l, and its expression is: Among them, H is the relative height difference between the front and rear points on the road centerline, and L is the horizontal distance between the front and rear points on the road centerline;

[0070] (c) Research shows that vehicle trajectories have the following three geometric characteristics:

[0071] 1) The trajectory is continuous and smooth;

[0072] 2) The trajectory has a continuous curvature, that is, there will be no sudden changes in the trajectory;

[0073] 3) The trajectory has a continuous curvature change rate, that is, there will be no sudden change in the trajectory change rate at any point on the driving trajectory.

[0074] The curvature of a straight line is 0, the curvature of a circular curve is 1 / r, and the curvature of a transition curve is a linear transition from 0 to 1 / r or from 1 / r to 0. Find all the road center points within the curvature search range A / 2, assuming A = 10m. These points form an arc, and the corresponding circular variance is: 2 +y 2 +ax+by+c=0, where x, y, z are the coordinates of the center point of the road, and a, b, c are the characteristic parameters of the circle. Use the least squares method to fit it, and solve a, b, c that minimizes the sum of squares of the residuals. Then you can calculate the radius r of the standard circle, and its reciprocal 1 / r is the curvature;

[0075] (d) Taking the center line of the road as the dividing point, the road surface is divided into two point sets on the left and right, and the least squares method is used to fit the straight line to obtain the slope, which is the cross slope.

[0076] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technology can understand and think of any changes or substitutions within the technical scope disclosed by the present invention, which should be included in the scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for extracting road geometry information based on laser point cloud, characterized by: The following steps are involved: (1) Perform radius filtering and grid downsampling on the point cloud collected by the vehicle-mounted laser scanning system to simplify the point cloud; (2) Extract the elevation features and local normal vector features of the simplified point cloud to distinguish ground points from non-ground points; The specific method of step (2) is: (a) Create a two-dimensional grid on the horizontal plane, project the simplified point cloud onto the horizontal plane and divide the point cloud into a number of squares according to the grid, restore the elevation information of the point cloud in each square, obtain a number of continuously distributed three-dimensional vertical columns, and select each vertical column one by one to perform the following steps (b) and (c); (b) Rearrange the point cloud in the vertical column in descending order of elevation value, and take the average elevation value of the 100 points with the lowest elevation value as the lowest elevation Z min , only keep the elevation value in Z min to Z min +□ ground Point cloud within the range, construct a point cloud set, where □ ground The maximum allowable range of ground point elevation change; (c) For each point cloud in the point cloud set constructed in (b), find the 20 nearest point clouds to construct a local point set, and standardize the point clouds in the local point set. Construct a covariance matrix for the standardized local point set, and solve the eigenvalues ​​and eigenvectors of the covariance matrix. The eigenvector corresponding to the minimum eigenvalue of the covariance matrix is ​​the normal vector of the point cloud. (d) Delete the point clouds whose normal vectors have an angle with the Z axis greater than 5°, and the remaining point clouds are ground points; (3) Based on the three-dimensional coordinates of the point cloud and the laser reflection intensity information, the region growing method is used on the ground points to connect the road surface points and remove the non-road surface parts of the ground points. After restoring the mistakenly deleted points, a complete road surface point cloud is obtained; In step (3), the specific method of using the region growing method on the ground points to connect the road points and remove the non-road parts of the ground points is: First, a point is randomly selected from the ground points as a seed point and the starting point of the region growth. Then, the neighboring points of this seed point are compared to see whether they meet the growth criteria. The growth criteria are: (a) The distance from the seed point is less than the set distance threshold; (b) The elevation difference of the seed point is less than the set elevation difference threshold; (c) The reflection intensity difference with the seed point is less than the set intensity difference threshold; Then, the neighboring points that meet the growth criteria are included in the region and used as new seed points of this region to continue growing outward until the growth ends; if the number of points in the region where the growth ends exceeds 50% of the total number of ground points, then the region is the road surface part of the ground points, otherwise the region is the non-road surface part of the ground points; select the non-road surface points of the non-road surface part one by one, compare their Euclidean distance and elevation difference with the road surface points, if the Euclidean distance between the non-road surface point and a road surface point is less than 1 / 2 of the set distance threshold and the elevation difference is less than the set elevation difference threshold, then restore the non-road surface point to a road surface point; (4) cutting a road cross section at the road surface point obtained in (3) according to the vehicle trajectory information; (5) Calculate the road width and longitudinal slope using the cross section, calculate the road curvature and transverse slope using the least squares method, and complete the extraction of road geometry information; The specific method of step (5) is: (a) Cross section CS corresponding to trajectory point i i In the figure, subtract the minimum abscissa from the maximum abscissa to obtain the cross section CS. i The width of the road surface at (b) Calculate the cross section CS corresponding to trajectory point i i The horizontal coordinate mean x_mean, the vertical coordinate mean y_mean and the elevation mean z_mean, then the point with (x_mean, y_mean, z_mean) as the coordinate That is the cross section CS i The road center point at the location; the road center point set obtained by calculating all trajectories constitutes the road center line, and the road center line elevation is used to calculate the road longitudinal slope l, which is expressed as: Among them, H is the relative height difference between the two points before and after the road centerline, and L is the horizontal distance between the two points before and after the road centerline; (c) For points on the road centerline Find all the road center points with it as the center and set A / 2 as the radius range. These road center points form an arc. Use the least squares method to fit the arc, solve the fitting parameters that minimize the sum of squared residuals, and then calculate the radius of the standard circle corresponding to the arc. The reciprocal of the radius is the curvature; where A is the search range when calculating the curvature; (d) Taking the center line of the road as the dividing point, the road surface is divided into two point sets on the left and right, and the least squares method is used to fit the straight line to obtain the slope, which is the cross slope.

2. The method for extracting road geometry information based on laser point cloud according to claim 1, characterized in that: The specific method of radius filtering and grid downsampling in step (1) is as follows: first, use the KD Tree method to establish a point cloud index; second, for each point cloud, if the total number of point clouds within the set radius neighborhood is less than the set threshold, the point cloud is deleted; finally, use a three-dimensional grid of a set size to divide the point cloud, and replace all the point clouds of this grid with the centroid point of each grid.

3. The method for extracting road geometry information based on laser point cloud according to claim 1, characterized in that: The specific method of step (4) is: The trajectory vector starting from the vehicle's trajectory point i It is expressed as: in and is the position vector of trajectory point i and trajectory point i+1; Calculate the normal vector of trajectory point i The expression is: Among them, v x 、v y 、v z are respectively characterized Projection length on the X, Y, and Z axes; Cut the road cross section along the normal vector direction of the trajectory point and rotate the cross section to the positive direction of the Y axis.

Images