A roadside laser radar deployment position assessment system and method

Abstract

The present invention discloses a roadside laser radar deployment position assessment system and method. The laser radar detection limit value is determined according to the selection of the laser radar, the perception space is set, and the point cloud data within the 360° range of the laser radar is obtained by scanning with the selected laser radar. The position of the point cloud data in the world coordinate system is calculated by the external parameter information of the selected laser radar installation. The complex laser radar installation position problem can be converted into a parameter calculation problem, and the evaluation of the laser radar installation position and angle and the selection of the optimal installation position can be realized. A metric based on each voxel point cloud is used to measure the amount of information perceived by the laser radar in the predefined perception space. The method of the present invention can simplify the original laser radar installation position assessment process, save the time consumption of the laser radar position assessment, be applicable to a variety of road structures, improve flexibility, and finally obtain an accurate laser radar installation position assessment index.

Description

Technical Field

[0001] The present invention belongs to the technical field of sensor deployment, and relates to a roadside laser radar deployment position assessment system and method. Background Art

[0002] In recent years, LiDAR has been widely used in the field of transportation due to its high resolution and ability to obtain 3D information of objects. The laser sensors installed on the roadside can provide global traffic flow information for intelligent transportation systems and autonomous vehicles with their "God's perspective" installation position, thus breaking through the bottleneck of limited vehicle perception distance. At present, most roadside sensing applications such as vehicle tracking, traffic monitoring, positioning, etc., the research on LiDAR perception methods is mostly focused on improving the performance of algorithms based on point cloud data. However, many studies have shown that the perception performance based on point cloud is not only affected by the algorithm, but also by the quality of the original point cloud data. The quality of the point cloud is directly affected by the installation position of LiDAR. However, determining the optimal installation position of LiDAR on the roadside is still a difficult problem. Due to the high cost of LiDAR, in the actual deployment process, it is necessary to test LiDARs of different locations and types one by one, and it is necessary to complete LiDAR deployment, data collection, model training, result evaluation and other tasks, which is extremely costly and consumes a lot of resources. Therefore, how to complete the quantitative evaluation of LiDAR configuration at a low cost in a short time is an urgent problem to be solved. Summary of the invention

[0003] The purpose of the present invention is to provide a roadside laser radar deployment position assessment system and method to overcome the problems of high cost and unreasonable layout caused by the existing laser radar installation relying on manual experience. The present invention can provide fast and effective roadside deployment guidance for the laser radar.

[0004] A method for evaluating a roadside laser radar deployment position includes the following steps:

[0005] S1, determine the selection and installation range of laser radar;

[0006] S2, determining the laser radar detection limit value according to the laser radar selection, setting the perception space based on the installation range and the laser radar detection limit value, and then voxelizing the set perception space;

[0007] S3, using the selected laser radar to scan and obtain point cloud data within the 360° range of the laser radar, establish a Cartesian coordinate system, convert the point cloud data into Cartesian coordinates, and then calculate the position of the point cloud data in the world coordinate system through the external parameter information installed by the selected laser radar;

[0008] S4, calculating the number of 3D point clouds acquired by all voxels in the perception space set in step S2;

[0009] S5, assigns a weight to each voxel in the perceptual space;

[0010] S6, calculating the evaluation index of the current lidar configuration using conditional entropy according to the number of 3D point clouds obtained by all voxels calculated in step S4 and the weight assigned to each voxel.

[0011] Preferably, the laser radar selection includes a specific model of the laser radar, and the installation range includes range information of a translation matrix and a rotation angle matrix of the laser radar relative to a world coordinate system.

[0012] Preferably, in step S2, the perception space is defined as a cuboid, and the definition includes the length l, width w, and height h of the cuboid; the resolution of the segmentation is γ, and the voxel segmentation in the perception space is expressed as:

[0013]

[0014] N represents the total number of voxels in the perceptual space.

[0015] Preferably, in step S3, the position of the point cloud data in the world coordinate system is calculated:

[0016] (3-1) Constructing the Cartesian coordinate system OXYZ: When the multi-line laser radar is located on a horizontal plane and is in a stationary state, the laser radar is taken as the center point, the vertical axis direction of the laser radar is taken as the Z axis, the ray direction parallel to the road is taken as the X axis, and the Y axis is determined by the Z axis and the X axis according to the right-hand screw rule;

[0017] (3-2) Calculate the coordinates of the point cloud data in the Cartesian coordinate system:

[0018]

[0019] The [x L ,y L , z L ] is the coordinate value of the point cloud data in the Cartesian coordinate system. The point where the laser beam touches the target object is called the return point. r represents the straight-line distance from the return point to the origin of the laser radar. The line connecting the detection point and the laser radar is used as the reference. θ represents the angle between the line and the z-axis. Represents the angle between the line and the x-axis.

[0020] Preferably, coordinate transformation of the point cloud data is performed according to the installation external parameters of the laser radar to obtain the data position in the world coordinate system, and the calculation formula is as follows:

[0021]

[0022] The [x w ,y w, z w ] is the coordinate value of the point cloud data in the world coordinate system, and R and T represent the rotation and translation matrices of the laser radar relative to the origin of the world coordinate system, respectively.

[0023] Preferably, in step S4, the number of point clouds in each voxel is calculated:

[0024] (4-1) According to formula (2), all points that can be generated by each beam in the region of interest are determined, where r is incremented by the resolution γ determined in S2, and the horizontal deflection angle is Determined by the horizontal resolution of the lidar;

[0025] (4-2) After calculating the Cartesian coordinates of the point cloud data, the coordinate system is transformed according to formula (3) to convert it to the world coordinate system. Each returned point is expressed as Among them, (x, y, z) represents its coordinates in the world coordinate system, and the set of all returned point coordinates is represented as p = {p 1, p 2, …,p T}, T represents the total number of all return points in the region of interest;

[0026] (4-3) According to the position of each voxel, the number of point clouds it includes is determined. For each voxel, the calculation formula is as follows:

[0027]

[0028] in, Represents the indicator function, which is used to indicate whether the returned point is located in a certain voxel. Combining formulas (2), (3), and (4), the number of point clouds corresponding to each voxel can be obtained.

[0029] Preferably, the weight calculation formula for each voxel is as follows:

[0030]

[0031] The w(v i ) represents the weight of each voxel, which can be specified according to different road structure types and attention requirements. Represents the weighted sum of all voxels.

[0032] Preferably, the concept of conditional entropy is introduced to evaluate the current lidar configuration, and for each voxel, the conditional entropy is calculated using the following formula:

[0033] Preferably, for the laser radar evaluation index under the current configuration, the higher the H value obtained, the better the current laser radar configuration is, and formula (7) is used for calculation:

[0034]

[0035] in represents the point cloud distribution within the voxel, and N represents the sum of all voxels.

[0036] A base road side laser radar deployment position assessment system, parameter input module, preprocessing module and assessment module;

[0037] The parameter input module is used to input the selection and installation range of the laser radar;

[0038] The preprocessing module is used to determine the detection limit of the laser radar according to the selected laser radar, set the perception space based on the installation range and the detection limit of the laser radar, and voxelize the set perception space; use the selected laser radar to scan and obtain the point cloud data within the 360° range of the laser radar, establish a Cartesian coordinate system, convert the point cloud data to Cartesian coordinates, and then calculate the position of the point cloud data in the world coordinate system through the external parameter information of the selected laser radar installation; calculate the number of 3D point clouds obtained by all voxels in the set perception space; and assign a weight to each voxel in the perception space;

[0039] The evaluation module uses conditional entropy to calculate the evaluation index output of the current lidar configuration based on the number of 3D point clouds obtained by all calculated voxels and the weight assigned to each voxel.

[0040] Compared with the prior art, the present invention has the following beneficial technical effects:

[0041] The present invention discloses a method for evaluating the deployment position of a roadside laser radar. The method determines the detection limit of the laser radar according to the selection of the laser radar, sets the perception space based on the installation range and the detection limit of the laser radar, and then voxelizes the set perception space. The selected laser radar is used to scan and obtain point cloud data within the 360° range of the laser radar, a Cartesian coordinate system is established, the point cloud data is converted into Cartesian coordinates, and the position of the point cloud data in the world coordinate system is calculated by the external parameter information of the selected laser radar installation. The method can convert the complex laser radar installation position problem into a parameter calculation problem, and can realize the evaluation of the installation position and angle of the laser radar and the selection of the optimal installation position. The method uses a metric based on each voxel point cloud to measure the amount of information perceived by the laser radar in the predefined perception space. The method of the present invention can simplify the original laser radar installation position evaluation process, namely sensor installation-data collection-model training-position evaluation, save the time consumption of the laser radar position evaluation, be applicable to a variety of road structures, improve flexibility, and finally obtain an accurate laser radar installation position evaluation index. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1This is a general framework diagram of the laser radar configuration evaluation in an embodiment of the present invention;

[0043] Figure 2 Schematic diagram of the laser radar spherical coordinate system in an embodiment of the present invention;

[0044] Figure 3 This is a diagram showing the comparison results of the deployment positions at different roll angles and heights in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0046] A method for evaluating the deployment position of a roadside laser radar according to the present invention comprises the following steps:

[0047] S1, determine the selection and installation range of the laser radar; the laser radar selection includes the specific model of the laser radar; the installation range includes the range information of the laser radar's translation matrix [x, y, z] and rotation angle matrix [roll, pitch, yaw] relative to the world coordinate system, and the origin of the world coordinate system can be defined by oneself, specifically the intersection point of the roadside facilities and the ground on which the laser radar depends;

[0048] S2, determining the laser radar detection limit value according to the laser radar selection, setting the perception space based on the installation range and the laser radar detection limit value, and then voxelizing the set perception space;

[0049] S3, using the selected laser radar to scan and obtain the point cloud data within the 360° range of the laser radar, establish the Cartesian coordinate system OXYZ, convert the point cloud data into Cartesian coordinates, and then calculate the position of the point cloud data in the world coordinate system through the external parameter information installed by the selected laser radar;

[0050] S4, calculating the number of 3D point clouds acquired by all voxels in the perception space set in step S2;

[0051] S5, assigns a weight to each voxel in the perception space to represent the importance of different voxels in the process of evaluating the LiDAR installation position;

[0052] S6, according to the number of 3D point clouds obtained by all voxels calculated in step S4 and the weight assigned to each voxel, the evaluation index of the current lidar configuration is calculated using conditional entropy, and the radar is configured and installed according to the evaluation index. The higher the evaluation index, the more suitable the corresponding configuration installation position. The present invention can realize the lidar configuration in a short time, which is simple and fast, and the conclusion is accurate.

[0053] In step S2, the perception space is defined as a cuboid, and the definition includes the length l, width w, and height h of the cuboid. The resolution of the segmentation is γ, so the voxel segmentation in the perception space is expressed as:

[0054]

[0055] N represents the total number of voxels in the perceptual space.

[0056] In step S3, the specific process of calculating the position of the point cloud data in the world coordinate system includes:

[0057] (3-1) Constructing the Cartesian coordinate system OXYZ: When the multi-line laser radar is located on a horizontal plane and is in a stationary state, the laser radar is taken as the center point, the vertical axis direction of the laser radar is taken as the Z axis, the ray direction parallel to the road is taken as the X axis, and the Y axis is determined by the Z axis and the X axis according to the right-hand screw rule;

[0058] (3-2) Calculate the coordinates of the point cloud data in the Cartesian coordinate system:

[0059]

[0060] The [x L ,y L , z L ] is the coordinate value of the point cloud data in the Cartesian coordinate system. The point where the laser beam touches the target object is called the return point. r represents the straight-line distance from the return point to the origin of the laser radar. The line connecting the detection point and the laser radar is used as the reference. θ represents the angle between the line and the z-axis. Represents the angle between the line and the x-axis.

[0061] (3-3) The coordinate transformation of the point cloud data is performed according to the installation external parameters of the laser radar to obtain the data position in the world coordinate system. The calculation formula is as follows:

[0062]

[0063] The [x w ,y w , z w ] is the coordinate value of the point cloud data in the world coordinate system, and R and T represent the rotation and translation matrices of the laser radar relative to the origin of the world coordinate system, respectively.

[0064] In step S4, the specific steps of calculating the number of point clouds in each voxel are as follows:

[0065] (4-1) According to formula (2), all points that can be generated by each beam in the region of interest are determined, where r is incremented by the resolution γ determined in S2, and the horizontal deflection angle is Determined by the horizontal resolution of the lidar;

[0066] (4-2) After calculating the Cartesian coordinates of the point cloud data, the coordinate system is transformed according to formula (3) to convert it to the world coordinate system. Each returned point is expressed as Among them, (x, y, z) represents its coordinates in the world coordinate system, and the set of all returned point coordinates can be represented as p = {p 1, p 2, …,p T}, T represents the total number of all return points in the region of interest;

[0067] (4-3) According to the position of each voxel, the number of point clouds it includes is determined. For each voxel, the calculation formula is as follows:

[0068]

[0069] in, Represents an indicator function, which indicates whether the returned point is located in a certain voxel. Combining formulas (2), (3), and (4), we can get the number of point clouds corresponding to each voxel.

[0070] In step S5, the weight calculation formula of each voxel is as follows:

[0071]

[0072] The w(v i ) represents the weight of each voxel, which can be specified according to different road structure types and attention requirements. Represents the weighted sum of all voxels.

[0073] In step S6, the steps of calculating the evaluation index of the current lidar configuration are as follows:

[0074] (6-1) The concept of conditional entropy is introduced to evaluate the current lidar configuration. For each voxel, its conditional entropy is calculated using the following formula:

[0075]

[0076] (6-2) The laser radar evaluation index under the current configuration is calculated using formula (7):

[0077]

[0078] in Represents the point cloud distribution within the voxel, and N represents the sum of all voxels. The higher the H value calculated in the end, the better the current LiDAR configuration.

[0079] The present invention provides a roadside laser radar installation evaluation method, which is applicable to various types of laser radars; this application takes the laser radar model Pandar64 as an example to analyze the method. The Pandar64 laser radar is composed of 64 laser emitting units in a row, and its vertical field of view angle is -25° to 15°. The beams are unevenly distributed in the vertical direction, the horizontal angular resolution is 0.2°, and it rotates horizontally at a frequency of 10 Hz per second. Environmental information within a 360° field of view of 100m around the laser radar as the origin can be provided. The data collected when the radar rotates one circle is designated as a frame, and the data output by the radar is expressed in spherical coordinates. The spherical coordinates are centered on the radar, and each radar has 5 parameters: vertical emission angle θ, horizontal emission angle The distance r from the point to the radar center, reflection intensity Intensity, and GPS time frame information.

[0080] like Figure 1 As shown in the figure, the specific process of the roadside lidar deployment location assessment method in this example is as follows:

[0081] S1, determine the type of LiDAR:

[0082] Specifically, determine the laser radar selection, including the manufacturer and specific model of the laser radar, such as the Pandar64 laser radar used in the implementation of the present invention. Different models of laser radars correspond to different field of view angles, resolutions, and number of beams. The installation allowable range refers to the external parameter installation range of the laser radar relative to the origin of the world coordinate system, including the limit range of translation distance and rotation angle.

[0083] S2, determine the detection limit of the lidar according to the selection of the lidar, predefine the perception space and voxelize it,

[0084] Specifically, in the example implementation of the present invention, the road structure is a straight road, and the perception space is defined as follows: l∈[-80m,80m], w∈[0,28m], h∈[0,3m]. Among them, l refers to the length of the perception space, w is the width of the perception space, and h refers to the height of the perception space. The defined perception space is voxelized with a segmentation resolution of 0.1m. Therefore, the voxel segmentation in the region of interest can be expressed as:

[0085]

[0086] Where N represents the total number of voxels in the perception space, and each voxel after splitting is represented in the perception space as, s i =[s x ,s y ,s z ], where i refers to the voxel number and [x, y, z] is the voxel's coordinate in the world coordinate system.

[0087] S3, establish a Cartesian coordinate system with the laser radar as the center and calculate the coordinates of the return point, which specifically includes the following steps:

[0088] S3.1, calculate the coordinates of the return point in the Cartesian coordinate system. LiDAR relies on emitting ultra-short laser pulses, projecting them onto objects for diffuse reflection, and then receiving the return beam. By measuring the flight time of the laser beam in the air, the distance between the target object and the sensor can be calculated. Therefore, the point where the laser beam touches the target object is called the return point.

[0089] Figure 2 This is a schematic diagram of the spherical coordinate system with the lidar as the origin. The deployment position is evaluated at different roll angles and heights. The evaluation results are as follows Figure 3 In the spherical coordinate system, the return point is defined by distance and two angles. The distance refers to the distance between the return point and the lidar origin. The two angles refer to the azimuth and polar angles, corresponding to Figure 2 The θ in The calculation formula for deriving the Cartesian coordinates of each returned point from the spherical coordinates is as follows:

[0090]

[0091] Among them, [x L ,y L , z L ] is the coordinate value of the point cloud data in the Cartesian coordinate system.

[0092] S3.2, determine the laser radar external parameter matrix: When determining the laser radar configuration in step S1, the input is the rotation angle information. Therefore, it is necessary to complete the transformation from the rotation angle information to the rotation angle matrix. The calculation formula is as follows:

[0093]

[0094] Roll represents the rotation angle of the LiDAR around the x-axis, pitch represents the rotation angle of the LiDAR around the y-axis, and yaw represents the rotation angle of the LiDAR around the z-axis.

[0095] Its translation matrix is:

[0096]

[0097] S3.3, determine the position conversion formula of the return point in the world coordinate system. According to the given translation matrix and rotation matrix information, the coordinate conversion of the point cloud data is performed to obtain the data position in the world coordinate system. The calculation formula is as follows:

[0098]

[0099] The [x w ,y w , z w ] is the coordinate value of the point cloud data in the world coordinate system, and R and T represent the rotation and translation matrices of the laser radar relative to the origin of the world coordinate system.

[0100] S3.4, calculate the coordinates of all returned points. As shown in formula (9), change r, θ, The coordinates of all returned points can be calculated. In the calculation process, r is kept constant, θ, By changing the vertical and horizontal resolutions of the laser radar, the coordinates of the return points of each beam at different horizontal angles and the same r value can be calculated. Finally, the calculated coordinate values ​​of the world coordinate system are normalized. The calculated return point coordinates are saved in the two-dimensional array point[i][j], where point[i] represents the number of all return points and point[i][j] represents the coordinates of each return point in the world coordinate system.

[0101] S4, calculating the number of point clouds contained in each voxel in the perception space, specifically including the following steps:

[0102] S4.1, initialize the current point coordinates, that is, make it equal to the translation value of the current installation position of the laser radar relative to the origin of the world coordinate system,

[0103]

[0104] S4.2, determine current z Is it less than the highest value of the predefined perception space? If so, jump to S4.3; otherwise, jump to S4.4.

[0105] S4.3, current reading x , current y , current z Are they all in the predefined perception space? If so, calculate x id ,y id ,z id , the calculation formula is as follows. If not, jump out of this loop and continue the calculation of the next return point.

[0106]

[0107] Among them, x_space_min, y_space_min, z_space_min represent the minimum values ​​of the defined perception space, and γ represents the resolution of voxelization of the perception space. id ,y id , z id By rounding, we can get the number of point clouds contained in a certain voxel. x , current y , current z The r value is increased, and the increasing formula is as follows:

[0108]

[0109] Jump back to the beginning of the loop, i.e. re-execute S4.3.

[0110] S4.4, if current z If the return point is greater than the maximum value defined by the sensing space, the distance d between the laser radar and the current point is calculated according to formula (16) to determine whether the current point has exceeded the effective detection range of the laser radar. If so, the loop is exited. If the return point is still within the effective detection range of the radar, the calculation is performed according to S4.3. The calculation formula of the distance d is as follows:

[0111]

[0112] S5, assign weights to each voxel in the perception space:

[0113]

[0114] In the example implementation process of the present invention, the weights of all voxels in the predefined perceptual space are 1, indicating that the voxels at all positions have the same attention.

[0115] S6, by introducing the conditional entropy formula, the evaluation index under the current lidar configuration can be calculated. The calculation formula is as follows:

[0116]

[0117] Among them, p(v i ) is the number of point clouds in each voxel calculated in S4.

[0118] H represents the final calculation result, that is, the lidar evaluation index under the current configuration. v represents the voxel, L0 represents the current laser configuration, and i represents the voxel number; Represents the evaluation index of a specific voxel in the current lidar configuration. represents the weight of a specific voxel, p(∑ i ) is the number of point clouds in each voxel calculated in S4. By weighting the evaluation values ​​of all voxels, the final evaluation index under the current LiDAR configuration can be obtained. The LiDAR configuration includes its rotation angle matrix (roll, pitch, yaw) and translation matrix (x, y, z) relative to the origin of the world coordinate system.

[0119] The present invention can transform the complex laser radar installation position problem into a parameter calculation problem, and can realize the evaluation of the laser radar installation position and angle and the selection of the optimal installation position. A metric based on per-voxel point cloud is used to measure the amount of information perceived by the laser radar in the predefined perception space. By using the method in the present invention, the original laser radar installation position evaluation process can be simplified, namely sensor installation-data collection-model training-position evaluation, which saves the time of laser radar position evaluation, is applicable to a variety of road structures, improves flexibility, and finally obtains an accurate laser radar installation position evaluation index.

Claims

1. A method for evaluating the deployment position of a roadside laser radar, characterized in that: The following steps are involved: S1, determine the selection and installation range of laser radar; S2, determining the laser radar detection limit value according to the laser radar selection, setting the perception space based on the installation range and the laser radar detection limit value, and then voxelizing the set perception space; S3, using the selected laser radar to scan and obtain point cloud data within the 360° range of the laser radar, establish a Cartesian coordinate system, convert the point cloud data into Cartesian coordinates, and then calculate the position of the point cloud data in the world coordinate system through the external parameter information installed by the selected laser radar; Establish a Cartesian coordinate system and convert the point cloud data into Cartesian coordinates: (3-1) Constructing the Cartesian coordinate system OXYZ: When the multi-line laser radar is located on a horizontal plane and is in a stationary state, the laser radar is taken as the center point, the vertical axis direction of the laser radar is taken as the Z axis, the ray direction parallel to the road is taken as the X axis, and the Y axis is determined by the Z axis and the X axis according to the right-hand screw rule; (3-2) Calculate the coordinates of the point cloud data in the Cartesian coordinate system: The [x L ,y L , z L ] is the coordinate value of the point cloud data in the Cartesian coordinate system. The point where the laser beam touches the target object is called the return point. r represents the straight-line distance from the return point to the origin of the laser radar. The line connecting the detection point and the laser radar is used as the reference. θ represents the angle between the line and the z-axis. Represents the angle between the connecting line and the x-axis; The position of the point cloud data in the world coordinate system is calculated through the external parameter information installed by the selected laser radar. The calculation formula is as follows: The [x w ,y w , z w ] is the coordinate value of the point cloud data in the world coordinate system, R and T represent the rotation and translation matrices of the laser radar relative to the origin of the world coordinate system respectively; S4, calculating the number of 3D point clouds obtained by all voxels in the perception space set in step S2; S5, assigns a weight to each voxel in the perceptual space; S6, according to the number of 3D point clouds obtained by all voxels calculated in step S4 and the weight assigned to each voxel, the evaluation index of the current lidar configuration is calculated using conditional entropy.

2. A method for evaluating the deployment position of a roadside laser radar according to claim 1, characterized in that: The laser radar selection includes the specific model of the laser radar, and the installation range includes the range information of the translation matrix and the rotation angle matrix of the laser radar relative to the world coordinate system.

3. The method for evaluating the deployment position of a roadside laser radar according to claim 1, characterized in that: In step S2, the perception space is defined as a cuboid, and the definition includes the length l, width w, and height h of the cuboid; the resolution of the segmentation is γ, and the voxel segmentation in the perception space is expressed as: N represents the total number of voxels in the perceptual space.

4. A method for evaluating the deployment position of a roadside laser radar according to claim 1, characterized in that: In step S4, the number of 3D point clouds acquired by all voxels in the perception space set in step S2 is calculated: (4-1) According to formula (2), all points that can be generated by each laser radar beam in the area of ​​interest are determined, where r is incremented by the resolution γ determined in S2, and the angle between the connecting line and the x-axis is Determined by the horizontal resolution of the lidar; (4-2) After calculating the Cartesian coordinates of the point cloud data, the coordinate system is transformed according to formula (3) to convert it to the world coordinate system. Each returned point is expressed as Among them, (x, y, z) represents its coordinates in the world coordinate system, and the set of all returned point coordinates is represented as p = {p1, p2, ..., p T }, T represents the total number of all return points in the region of interest; (4-3) According to the position of each voxel, the number of 3D point clouds it contains is determined. For each voxel, the calculation formula is as follows: in, represents the indicator function, which is used to indicate whether the returned point is located in a certain voxel. Combining formulas (2), (3), and (4), we can get the number of 3D point clouds corresponding to each voxel; p(v i ) is the number of point clouds in each voxel.

5. The method for evaluating the deployment position of a roadside laser radar according to claim 1, characterized in that: The weight calculation formula for each voxel is as follows: The w(v i ) represents the weight of each voxel, which can be specified according to different road structure types and attention requirements. Represents the weighted sum of all voxels.

6. A method for evaluating the deployment position of a roadside laser radar according to claim 1, characterized in that: The concept of conditional entropy is introduced to evaluate the current lidar configuration. For each voxel, its conditional entropy is calculated using the following formula: In the formula, w(v i ) is the weight of each voxel; p(v i ) is the number of point clouds in each voxel; L0 represents the current laser configuration, i represents the voxel number, and v represents the voxel.

7. A method for evaluating the deployment position of a roadside laser radar according to claim 1, characterized in that: For the laser radar evaluation index under the current configuration, the higher the H value obtained, the better the current laser radar configuration is. Formula (7) is used for calculation: in represents the point cloud distribution within the voxel, and N represents the sum of all voxels.

8. A roadside laser radar deployment position assessment system based on the method of claim 1, characterized in that: Parameter input module, preprocessing module and evaluation module; The parameter input module is used to input the selection and installation range of the laser radar; The preprocessing module is used to determine the detection limit of the laser radar according to the selected laser radar, set the perception space based on the installation range and the detection limit of the laser radar, and voxelize the set perception space; use the selected laser radar to scan and obtain the point cloud data within the 360° range of the laser radar, establish a Cartesian coordinate system, convert the point cloud data to Cartesian coordinates, and then calculate the position of the point cloud data in the world coordinate system through the external parameter information of the selected laser radar installation; calculate the number of 3D point clouds obtained by all voxels in the set perception space; Assign a weight to each voxel in the perception space; Establish a Cartesian coordinate system and convert the point cloud data into Cartesian coordinates: (3-1) Constructing the Cartesian coordinate system OXYZ: When the multi-line laser radar is located on a horizontal plane and is in a stationary state, the laser radar is taken as the center point, the vertical axis direction of the laser radar is taken as the Z axis, the ray direction parallel to the road is taken as the X axis, and the Y axis is determined by the Z axis and the X axis according to the right-hand screw rule; (3-2) Calculate the coordinates of the point cloud data in the Cartesian coordinate system: The [x L ,y L , z L ] is the coordinate value of the point cloud data in the Cartesian coordinate system. The point where the laser beam touches the target object is called the return point. r represents the straight-line distance from the return point to the origin of the laser radar. The line connecting the detection point and the laser radar is used as the reference. θ represents the angle between the line and the z-axis. Represents the angle between the connecting line and the x-axis; The position of the point cloud data in the world coordinate system is calculated through the external parameter information installed by the selected laser radar. The calculation formula is as follows: The [x w ,y w , z w ] is the coordinate value of the point cloud data in the world coordinate system, R and T represent the rotation and translation matrices of the laser radar relative to the origin of the world coordinate system respectively; The evaluation module uses conditional entropy to calculate the evaluation index output of the current lidar configuration based on the number of 3D point clouds obtained by all calculated voxels and the weight assigned to each voxel.

Images