A laser radar ground extraction method and device

Abstract

The present invention relates to a laser radar ground extraction method and device. First, the laser radar is calibrated so that the Z-axis values ​​of the collected point cloud data are not less than 0. Secondly, the laser point cloud data is divided into multiple concentric circle areas according to the vehicle body pitch angle and the laser point cloud distance. Then, the concentric circle areas are divided into multiple sectors. For each sector, the laser point cloud data is sorted according to the Z-axis. The laser point cloud data with the smaller Z-axis value is used for plane fitting. The laser point cloud data is initially divided into a ground point set and a non-ground point set, and the ground point set obtained by the initial division is recorded as the initial ground point set. Finally, the initial ground point set of each sector is respectively subjected to normal vector filtering and vertical elevation filtering, and the initial ground point set is judged based on the filtering results to obtain a final ground point set. The present invention has a good recognition effect on roads and fences, road junctions, and slopes less than 45°.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a laser radar ground extraction method and device. Background Art

[0002] During the LiDAR target recognition process, the ground needs to be extracted to distinguish between ground and non-ground. Currently, the following methods are commonly used:

[0003] 1. Use RANSAN to perform plane fitting, and then determine whether the distance z from the point cloud to the plane is within the threshold to determine whether it is the ground.

[0004] 2. Project the laser point cloud into the polar coordinate system to form the DZ coordinate system, and then use the Gaussian or linear fitting method to distinguish between the ground and non-ground.

[0005] 3. Order the disordered point cloud and use the characteristics of the lidar itself to detect whether the angle between the upper and lower laser point clouds meets the vertical angle of the lidar to distinguish between ground and non-ground.

[0006] These methods can distinguish the ground in specific scenarios, but in some cases, false detection may occur, such as in slopes, road squats, weeds and other scenarios, and the ground cannot be distinguished correctly and effectively. Summary of the Invention

[0007] In response to the technical problems existing in the prior art, the present invention provides a lidar ground extraction method and device, which is optimized for scenes such as ramps, road dips, and weeds. The point cloud area is divided according to the different pitch angles of the vehicle body, and filtered according to weight factors such as normal vectors and elevation. It has a good recognition effect on the connection between roads and fences, road dips, and slopes less than 45°.

[0008] The technical solution of the present invention to solve the above technical problems is as follows:

[0009] In a first aspect, the present invention provides a laser radar ground extraction method, comprising:

[0010] Calibrate the LiDAR so that the Z-axis values ​​of the collected point cloud data are not less than 0;

[0011] According to the vehicle body pitch angle and laser point cloud distance, the laser point cloud data is divided into multiple concentric circle areas;

[0012] Divide the concentric circle area into multiple sectors, sort the laser point cloud data according to the Z axis for each sector, take the laser point cloud data with smaller Z axis value for plane fitting, and preliminarily divide the laser point cloud data into a ground point set and a non-ground point set, and record the ground point set obtained by the initial division as the initial ground point set;

[0013] The initial ground point set of each sector is subjected to normal vector filtering and vertical elevation filtering respectively, and the initial ground point set is judged according to the filtering results to obtain the final ground point set.

[0014] Furthermore, according to the vehicle body pitch angle and laser point cloud distance, the laser point cloud data is divided into multiple concentric circle areas, including:

[0015] Note: H1 is the vertical height of the sensor; R1 is the radius of the set ring; α is the angle between the laser beam and the horizontal plane, that is, the pitch angle of the vehicle body; R2 is the actual distance the laser hits the ramp when the slope is β, which is also the actual concentric circle radius; then:

[0016]

[0017] According to empirical values, when the vehicle body pitch angle α is 0°, the laser point cloud data is divided into multiple concentric circle areas, and the actual distance the laser hits the ramp, i.e. the actual concentric circle radius, is calculated according to the above formula when the slope of the ramp is β.

[0018] Furthermore, for each sector, the laser point cloud data is sorted according to the Z axis, and the laser point cloud data with smaller Z axis values ​​is used for plane fitting, and the laser point cloud data is preliminarily divided into ground point sets and non-ground point sets, including:

[0019] For each sector, the laser point cloud data is sorted according to the Z axis, and the 20% laser point cloud data with the smallest Z axis value is obtained and plane fitting is performed;

[0020] Use PCA principal component analysis to obtain the distance D from the point cloud to the fitting plane;

[0021] Set the distance threshold th_d, classify the point clouds with a distance D less than the distance threshold th_d as ground point sets, and classify the other point clouds as non-ground point sets.

[0022] Furthermore, normal vector filtering and vertical elevation filtering are performed on the initial ground point set of each sector, and the initial ground point set is judged according to the filtering results to obtain the final ground point set, including:

[0023] For each sector, obtain the normal vector of the fitting plane of the initial ground point set. If the direction of the normal vector is greater than the slope β, calculate the normal vector weight factor δ of the initial ground point set of the current sector according to the following formula;

[0024] Normal represents the direction of the normal vector;

[0025] Set the vertical height threshold th_elevation of each concentric circle area respectively, and obtain the vertical elevation mean of each sector in each concentric circle area n is the sector number; the elevation weight factor of the initial ground point set of the current sector is calculated according to the following formula

[0026]

[0027] Calculate the probability f(x|β) that the initial ground point set of the current sector is the final ground point set,

[0028] When f(x|β)>0.5, the initial ground point set of the current sector is the final ground point set.

[0029] Furthermore, based on experience, when the vehicle body pitch angle α is 0°, the circle radius R1 is taken as 2.7m, 12.3m, 23.2m, and 47.5m respectively, dividing the laser point cloud data into four concentric circle areas.

[0030] Furthermore, the concentric circle areas are divided into a plurality of sectors, including: dividing the four concentric circle areas into 16, 32, 32, and 16 sectors respectively.

[0031] Furthermore, the vertical height thresholds th_elevation of the concentric circle areas are 3.25, 5.2, 6.5, and 8.9 from the inside to the outside.

[0032] In a second aspect, the present invention provides a laser radar ground extraction device, comprising:

[0033] The calibration module is used to calibrate the laser radar so that the Z-axis value of the collected point cloud data is not less than 0;

[0034] The first division module is used to divide the laser point cloud data into multiple concentric circle areas according to the vehicle body pitch angle and the laser point cloud distance;

[0035] The second division module is used to divide the concentric circle area into multiple sectors, sort the laser point cloud data according to the Z axis for each sector, take the laser point cloud data with a smaller Z axis value for plane fitting, and preliminarily divide the laser point cloud data into a ground point set and a non-ground point set, and record the ground point set obtained by the initial division as the initial ground point set;

[0036] The filtering and judgment module is used to perform normal vector filtering and vertical elevation filtering on the initial ground point set of each sector, and judge the initial ground point set based on the filtering results to obtain the final ground point set.

[0037] In a third aspect, the present invention provides an electronic device, comprising:

[0038] Memory for storing computer software programs;

[0039] The processor is used to read and execute the computer software program, thereby implementing the laser radar ground extraction method described in the first aspect of the present invention.

[0040] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium, in which a computer software program for implementing a laser radar ground extraction method described in the first aspect of the present invention is stored.

[0041] The present invention has the following advantages: It resolves point cloud density issues by dividing the point cloud into different regions. Within the region division, sector sizes are differentiated based on different slopes, enhancing algorithm adaptability. It also uses PCA master-slave analysis to improve ground fitting speed. It also uses weighting factors such as normal vectors and elevation to address mismatching issues. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic diagram of a method flow chart provided in an embodiment of the present invention;

[0043] Figure 2 A schematic diagram of the structure of a device provided in an embodiment of the present invention;

[0044] Figure 3 A schematic diagram of an electronic device according to an embodiment of the present invention;

[0045] Figure 4 A schematic diagram of an embodiment of a computer-readable storage medium provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0047] like Figure 1 As shown, an embodiment of the present invention provides a laser radar ground extraction method, comprising the following steps:

[0048] S1: Calibrate the LiDAR so that the Z-axis values ​​of the collected point cloud data are not less than 0.

[0049] When we get the lidar data, many of the Z-axis values ​​of the lidar hitting the ground are negative. We calibrate the lidar so that the Z-axis data of the laser point cloud are greater than or equal to 0, thus completing the lidar calibration.

[0050] S2, divides the laser point cloud data into multiple concentric circle areas according to the vehicle body pitch angle and the laser point cloud distance.

[0051] Note: H1 is the vertical height of the sensor; R1 is the radius of the set ring; α is the angle between the laser beam and the horizontal plane, that is, the pitch angle of the vehicle body; R2 is the actual distance the laser hits the ramp when the slope is β, which is also the actual concentric circle radius; then:

[0052]

[0053] According to empirical values, when the vehicle body pitch angle α is 0°, the laser point cloud data is divided into multiple concentric circle areas, and the actual distance the laser hits the ramp, i.e. the actual concentric circle radius, is calculated according to the above formula when the slope of the ramp is β.

[0054] We divide the laser point cloud into regions according to distance and use IMU data to obtain the pitch angle of the vehicle body. When the pitch angle is around 0°, four concentric circles are divided at 2.7m, 12.3m, 23.2m, and 47.5m (region division is performed according to formula (1) at 10°, 20°, 30°, 40°, and 45°).

[0055] S3, dividing the concentric circle area into multiple sectors, sorting the laser point cloud data according to the Z axis for each sector, taking the laser point cloud data with a smaller Z axis value for plane fitting, and preliminarily dividing the laser point cloud data into a ground point set and a non-ground point set, and recording the ground point set obtained by the initial division as the initial ground point set.

[0056] The four concentric circles are further divided into different numbers of sectors: 16, 32, 32, and 16. Sector segmentation is used to solve the point cloud density problem. The point cloud density is higher near, so the blocks are smaller. The point cloud density is lower far, so the blocks are larger.

[0057] For each sector, we sort the point cloud along the Z axis and obtain the 20% smallest laser point cloud. This point cloud is then used for plane fitting. We perform PCA principal component analysis to determine the distance d between the point cloud and the fitting plane. If the distance d is less than a threshold th_d, the point cloud is included in the ground point cloud set for the next plane fitting. If it is greater than the threshold, the point cloud is considered a non-ground point cloud. After three fitting cycles, the point cloud and non-point cloud are distinguished.

[0058] S4, performing normal vector filtering and vertical elevation filtering on the initial ground point set of each sector respectively, and judging the initial ground point set according to the filtering results to obtain the final ground point set.

[0059] For each sector, obtain the normal vector of the fitting plane of the initial ground point set. If the direction of the normal vector is greater than the slope β, calculate the normal vector weight factor δ of the initial ground point set of the current sector according to the following formula;

[0060] Normal represents the direction of the normal vector;

[0061] Set the vertical height threshold th_elevation of each concentric circle area respectively, and obtain the vertical elevation mean of each sector in each concentric circle area n is the sector number; the elevation weight factor of the initial ground point set of the current sector is calculated according to the following formula

[0062]

[0063] Calculate the probability f(x|β) that the initial ground point set of the current sector is the final ground point set,

[0064] When f(x|β)>0.5, the initial ground point set of the current sector is the final ground point set.

[0065] The vertical height thresholds th_elevation of the concentric circle areas are 3.25, 5.2, 6.5, and 8.9 from the inner to the outer. The vertical elevation mean refers to the average vertical elevation of all points in a sector.

[0066] like Figure 2 As shown, an embodiment of the present invention further provides a laser radar ground extraction device, comprising:

[0067] The calibration module is used to calibrate the laser radar so that the Z-axis value of the collected point cloud data is not less than 0;

[0068] The first division module is used to divide the laser point cloud data into multiple concentric circle areas according to the vehicle body pitch angle and the laser point cloud distance;

[0069] The second division module is used to divide the concentric circle area into multiple sectors, sort the laser point cloud data according to the Z axis for each sector, take the laser point cloud data with a smaller Z axis value for plane fitting, and preliminarily divide the laser point cloud data into a ground point set and a non-ground point set, and record the ground point set obtained by the initial division as the initial ground point set;

[0070] The filtering and judgment module is used to perform normal vector filtering and vertical elevation filtering on the initial ground point set of each sector, and judge the initial ground point set based on the filtering results to obtain the final ground point set.

[0071] See also Figure 3 , Figure 3 Schematic diagram of an embodiment of an electronic device provided by an embodiment of the present invention. Figure 3As shown, an embodiment of the present invention provides an electronic device 500, including a memory 510, a processor 520, and a computer program 511 stored in the memory 520 and executable on the processor 520. When the processor 520 executes the computer program 511, the following steps are implemented:

[0072] Calibrate the LiDAR so that the Z-axis values ​​of the collected point cloud data are not less than 0;

[0073] According to the vehicle body pitch angle and laser point cloud distance, the laser point cloud data is divided into multiple concentric circle areas;

[0074] Divide the concentric circle area into multiple sectors, sort the laser point cloud data according to the Z axis for each sector, take the laser point cloud data with smaller Z axis value for plane fitting, and preliminarily divide the laser point cloud data into a ground point set and a non-ground point set, and record the ground point set obtained by the initial division as the initial ground point set;

[0075] The initial ground point set of each sector is subjected to normal vector filtering and vertical elevation filtering respectively, and the initial ground point set is judged according to the filtering results to obtain the final ground point set.

[0076] See also Figure 4 , Figure 4 Schematic diagram of an embodiment of a computer-readable storage medium provided in an embodiment of the present invention. Figure 4 As shown, this embodiment provides a computer-readable storage medium 600 on which a computer program 611 is stored. When the computer program 611 is executed by a processor, the following steps are implemented:

[0077] Calibrate the LiDAR so that the Z-axis values ​​of the collected point cloud data are not less than 0;

[0078] According to the vehicle body pitch angle and laser point cloud distance, the laser point cloud data is divided into multiple concentric circle areas;

[0079] Divide the concentric circle area into multiple sectors, sort the laser point cloud data according to the Z axis for each sector, take the laser point cloud data with smaller Z axis value for plane fitting, and preliminarily divide the laser point cloud data into a ground point set and a non-ground point set, and record the ground point set obtained by the initial division as the initial ground point set;

[0080] The initial ground point set of each sector is subjected to normal vector filtering and vertical elevation filtering respectively, and the initial ground point set is judged according to the filtering results to obtain the final ground point set.

[0081] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0082] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0084] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0086] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0087] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A laser radar ground extraction method, characterized in that: include: Calibrate the LiDAR so that the Z-axis values ​​of the collected point cloud data are not less than 0; According to the vehicle body pitch angle and laser point cloud distance, the laser point cloud data is divided into multiple concentric circle areas; Divide the concentric circle area into multiple sectors, sort the laser point cloud data according to the Z axis for each sector, take the laser point cloud data with smaller Z axis value for plane fitting, and preliminarily divide the laser point cloud data into a ground point set and a non-ground point set, and record the ground point set obtained by the initial division as the initial ground point set; The initial ground point set of each sector is filtered by normal vector and vertical elevation respectively, and the initial ground point set is judged according to the filtering results to obtain the final ground point set; The initial ground point set of each sector is filtered by normal vector and vertical elevation respectively, and the initial ground point set is judged according to the filtering results to obtain the final ground point set, including: For each sector, obtain the normal vector of the fitting plane of the initial ground point set. If the direction of the normal vector is greater than the slope , then the normal vector weight factor of the initial ground point set of the current sector is calculated according to the following formula ; , normal represents the direction of the normal vector; Set the vertical height threshold th_elevation of each concentric circle area respectively, and obtain the vertical elevation mean of each sector in each concentric circle area , is the sector number; the elevation weight factor of the initial ground point set of the current sector is calculated according to the following formula ; ； Calculate the probability that the initial ground point set of the current sector is the final ground point set , ; is the radius of the ring to be set; Indicates that when the slope of the ramp is When , the actual distance the laser hits the ramp is also the actual concentric circle radius; when , then the initial ground point set of the current sector is the final ground point set.

2. The method according to claim 1, characterized in that According to the vehicle body pitch angle and laser point cloud distance, the laser point cloud data is divided into multiple concentric circle areas, including: remember: Indicates the vertical height of the sensor; is the radius of the ring to be set; Indicates the angle between the laser beam and the horizontal plane, that is, the vehicle body pitch angle; Indicates that when the slope of the ramp is When , the actual distance the laser hits the ramp is also the actual concentric circle radius; then: ， According to the experience, the vehicle pitch angle When the slope is 0°, the laser point cloud data is divided into multiple concentric circle areas, and the slope of the ramp is calculated according to the above formula. When the laser hits the ramp, the actual distance is the actual concentric circle radius.

3. The method according to claim 2, characterized in that For each sector, the laser point cloud data is sorted according to the Z axis, and the laser point cloud data with the smaller Z axis value is used for plane fitting. The laser point cloud data is preliminarily divided into ground point set and non-ground point set, including: For each sector, the laser point cloud data is sorted according to the Z axis, and the 20% laser point cloud data with the smallest Z axis value is obtained and plane fitting is performed; Use PCA principal component analysis to obtain the distance D from the point cloud to the fitting plane; Set the distance threshold th_d, classify the point clouds with a distance D less than the distance threshold th_d as ground point sets, and classify the other point clouds as non-ground point sets.

4. The method according to claim 3, characterized in that According to experience, the vehicle pitch angle When it is 0°, the radius of the ring The laser point cloud data is divided into four concentric circle areas at 2.7m, 12.3m, 23.2m, and 47.5m respectively.

5. The method according to claim 4, characterized in that Dividing the concentric circle areas into a plurality of sectors includes: dividing the four concentric circle areas into 16, 32, 32, and 16 sectors respectively.

6. The method according to claim 5, characterized in that The vertical height thresholds th_elevation of the concentric circle areas are 3.25, 5.2, 6.5, and 8.9 from the inner to the outer.

7. A laser radar ground extraction device, characterized in that: include: The calibration module is used to calibrate the laser radar so that the Z-axis value of the collected point cloud data is not less than 0; The first division module is used to divide the laser point cloud data into multiple concentric circle areas according to the vehicle body pitch angle and the laser point cloud distance; The second division module is used to divide the concentric circle area into multiple sectors, sort the laser point cloud data according to the Z axis for each sector, take the laser point cloud data with a smaller Z axis value for plane fitting, and preliminarily divide the laser point cloud data into a ground point set and a non-ground point set, and record the ground point set obtained by the initial division as the initial ground point set; The filtering and judgment module is used to perform normal vector filtering and vertical elevation filtering on the initial ground point set of each sector, and judge the initial ground point set based on the filtering results to obtain the final ground point set; The initial ground point set of each sector is filtered by normal vector and vertical elevation respectively, and the initial ground point set is judged according to the filtering results to obtain the final ground point set, including: For each sector, obtain the normal vector of the fitting plane of the initial ground point set. If the direction of the normal vector is greater than the slope , then the normal vector weight factor of the initial ground point set of the current sector is calculated according to the following formula ; , normal represents the direction of the normal vector; Set the vertical height threshold th_elevation of each concentric circle area respectively, and obtain the vertical elevation mean of each sector in each concentric circle area , is the sector number; the elevation weight factor of the initial ground point set of the current sector is calculated according to the following formula ; ； Calculate the probability that the initial ground point set of the current sector is the final ground point set , ; is the radius of the ring to be set; Indicates that when the slope of the ramp is When , the actual distance the laser hits the ramp is also the actual concentric circle radius; when , then the initial ground point set of the current sector is the final ground point set.

8. An electronic device, characterized in that: include: Memory for storing computer software programs; A processor is used to read and execute the computer software program, thereby implementing a laser radar ground extraction method as described in any one of claims 1-6.

9. A non-transitory computer-readable storage medium, characterized in that The storage medium stores a computer software program for implementing a laser radar ground extraction method as described in any one of claims 1-6.

Images