Bit-based laser radar point cloud sparsification method and system, and storage medium

Abstract

The invention relates to a bit-based laser radar point cloud rarefaction method and system and a storage medium, and the method comprises the steps: M1, initializing a grid, and obtaining the data information of a laser radar point cloud of a current frame; m2, traversing the data information of the laser radar point cloud of the current frame by using a bit-based laser radar point cloud sparsification algorithm according to the initialized grid, and performing sparsification processing on the laser radar point cloud to obtain the data information of the radar point cloud of the current frame after sparsification; and M3, based on the data information of the rarefied radar point cloud of the current frame, optimizing the rarefied radar point cloud by adopting an improved transient search optimization algorithm to obtain the final data information of the radar point cloud of the current frame. According to the method, the point cloud processing time and the computing platform resource occupation can be obviously reduced, and the point cloud sparsification time and the resource occupation can be greatly reduced while the point cloud quality is maintained.

Description

Technical Field

[0001] This invention relates to the field of lidar point cloud processing technology, and in particular to a bit-based lidar point cloud sparsification method, system, and storage medium. Background Technology

[0002] LiDAR is one of the core sensors in autonomous driving perception technology. Its precise spatial measurement capabilities enable autonomous driving systems to acquire more intuitive and detailed environmental information. However, the massive amount of LiDAR point cloud data and its uneven density result in high time costs for its application modules and affect algorithm stability. To ensure the real-time performance and stability of the autonomous driving system, it is necessary to perform uniform sparsification on the dense original point cloud.

[0003] Existing point cloud sparsification algorithms typically use a gridded approach to represent the point cloud surrounding the vehicle. The core idea is to initialize a grid; then traverse the point cloud, calculating the grid coordinates of each point (let's say the i-th grid) and storing that point in the corresponding point cloud queue; finally, after evaluating all point clouds, the center point of all points in each grid is calculated as the point cloud for that grid. Because existing point cloud sparsification algorithms directly evaluate the point cloud after calculating its grid coordinates, and project all point clouds onto the grid while calculating the center points within each grid, this process is time-consuming and consumes significant computational resources, which is highly detrimental to real-time autonomous driving systems. Summary of the Invention

[0004] In view of the above problems, the present invention provides a bit-based method, system and storage medium for sparsification of LiDAR point clouds, which can not only significantly reduce the time and computing platform resources occupied by point cloud processing, but also greatly reduce the time and resources occupied by point cloud sparsification while maintaining the quality of point cloud.

[0005] To achieve the above and other related objectives, the present invention provides the following technical solution: A bit-based method for sparsifying point clouds of lidar, the method comprising: M1. Initialize the grid and acquire the LiDAR point cloud data information for the current frame; M2. Based on the initial grid, a bit-based LiDAR point cloud sparsification algorithm is used to traverse the data information of the LiDAR point cloud in the current frame, and the LiDAR point cloud is sparsified to obtain the data information of the LiDAR point cloud after sparsification in the current frame. M3. Based on the data information of the sparsed radar point cloud of the current frame, an improved transient search optimization algorithm is used to optimize the sparsed radar point cloud to obtain the final radar point cloud data information of the current frame.

[0006] Furthermore, in step M2, the step of using a bit-based lidar point cloud sparsification algorithm to traverse the lidar point cloud data information of the current frame and perform sparsification processing on the lidar point cloud includes: M21. Project the data information of the current frame's lidar point cloud onto the initialized grid, and calculate the i-th lidar point cloud (pt). ix , pt iy , pt iz The corresponding raster coordinates (x) Axis , y Axis , z Axis ), , , , Where (minX,minY,minZ) are the lower boundary points of the initialized raster; M22. Determine whether the current grid contains points from the lidar point cloud based on bit method. If it does, return to step M21; otherwise, proceed to step M23. M23. Store the i-th point into the sparse point cloud queue to obtain the data information of the sparse radar point cloud of the current frame.

[0007] Furthermore, in step M22, the bit-based method of determining whether the current grid contains points from the lidar point cloud includes: M221. Based on the i-th lidar point cloud (pt) ix , pt iy , pt iz The corresponding raster coordinates (x) Axis ,y Axis , z Axis ), calculate the raster index id x , , Where (maxX,maxY,maxZ) are the upper boundary points of the initialized raster; M222. Calculate the byte number charNum where the bit is located, the bit offset charBit, the bit mask moveBit, and the flag bit hasPt indicating whether there is a point in the grid. M223. Determine whether the grid cell containing the current point has been projected into a point cloud. If hasPt = 0, it means that the current grid cell has not been projected into a point cloud. Then, store the point in the sparsed point cloud list. Otherwise, hasPt = 1, the current grid cell already contains a point, and perform grid projection on the (i+1)th point.

[0008] Furthermore, the byte number charNum where the bit is located is charNum=id. x / 8, the bit offset charBit is charBit=id x %8, the bit mask moveBit is moveBit=(1< <charBit)。

[0009] Furthermore, the flag bit hasPt of the point is, hasPt=havePoint[charNum]&moveBit.

[0010] Furthermore, in step M3, the optimization of the sparsed radar point cloud using the improved transient search optimization algorithm includes: M31. Based on the data information of the radar point cloud after the current frame is sparsified, the population is initialized, the population parameters and the maximum number of iterations L are determined, and the data information of the initialized population is obtained. M32. Based on the data information of the initialized population, update the thermal resistance coefficient C of the population. l and T; M33. Update the population position W(X,C) l ,T); M34. Evaluate fitness and update optimal position; M35. Determine if the maximum number of iterations L is satisfied. If it is, output the optimal position and the global optimal solution, which are the final radar point cloud data information for the current frame. If it is not satisfied, return to step M32.

[0011] Furthermore, the thermal resistance coefficient C of the population l And T is, , , Where z is the attenuation coefficient variable changing from 2 to 0, k is a real number, and r1, r2 and r3 are random numbers in [0,1].

[0012] Furthermore, the location of the population W(X,C) l ,T) is, , Where X represents the individual parameters of the initialized population, and e represents the natural base.

[0013] To achieve the above and other related objectives, the present invention also provides a bit-based lidar point cloud sparsification system, including a computer device programmed or configured to perform the steps of the bit-based lidar point cloud sparsification method.

[0014] To achieve the above and other related objectives, the present invention also provides a computer-readable storage medium storing a computer program programmed or configured to perform the bit-based lidar point cloud sparsification method described above.

[0015] The present invention has the following positive effects: This invention significantly reduces point cloud processing time and computing platform resource consumption by performing point cloud sparsification based on bits. Furthermore, by using the first point as the point cloud within a grid, it maintains point cloud quality while greatly reducing sparsification time and resource consumption. This reduction in processing time and resource consumption provides subsequent perception algorithms with sufficient time and resources, enabling real-time and stable environmental information output, thereby improving the real-time performance and robustness of the autonomous driving system. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a flowchart illustrating the bit-based point cloud sparsification algorithm of the present invention. Figure 3 This is a flowchart illustrating the improved transient search optimization algorithm of the present invention. Detailed Implementation

[0017] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0018] Example 1: As Figure 1 As shown, a bit-based method for sparsifying LiDAR point clouds is described, the method comprising: M1. Initialize the grid and acquire the LiDAR point cloud data information for the current frame; M2. Based on the initial grid, a bit-based LiDAR point cloud sparsification algorithm is used to traverse the data information of the LiDAR point cloud in the current frame, and the LiDAR point cloud is sparsified to obtain the data information of the LiDAR point cloud after sparsification in the current frame. M3. Based on the data information of the sparsed radar point cloud of the current frame, an improved transient search optimization algorithm is used to optimize the sparsed radar point cloud to obtain the final radar point cloud data information of the current frame.

[0019] In this embodiment, as Figure 2 As shown, in step M2, the step of using a bit-based lidar point cloud sparsification algorithm to traverse the lidar point cloud data information of the current frame and perform sparsification processing on the lidar point cloud includes: M21. Project the data information of the current frame's lidar point cloud onto the initialized grid, and calculate the i-th lidar point cloud (pt). ix , pt iy , pt iz The corresponding raster coordinates (x) Axis , y Axis , z Axis ), , , , Where (minX,minY,minZ) are the lower boundary points of the initialized raster; M22. Determine whether the current grid contains points from the lidar point cloud based on bit method. If it does, return to step M21; otherwise, proceed to step M23. M23. Store the i-th point into the sparse point cloud queue to obtain the data information of the sparse radar point cloud of the current frame.

[0020] In this embodiment, in step M22, determining whether the current grid contains points from the lidar point cloud based on bit-based methods includes: M221. Based on the i-th lidar point cloud (pt) ix , pt iy , pt iz The corresponding raster coordinates (x) Axis ,y Axis , z Axis ), calculate the raster index id x , , Where (maxX,maxY,maxZ) are the upper boundary points of the initialized raster; M222. Calculate the byte number charNum where the bit is located, the bit offset charBit, the bit mask moveBit, and the flag bit hasPt indicating whether there is a point in the grid. M223. Determine whether the grid cell containing the current point has been projected into a point cloud. If hasPt = 0, it means that the current grid cell has not been projected into a point cloud. Then, store the point in the sparsed point cloud list. Otherwise, hasPt = 1, the current grid cell already contains a point, and perform grid projection on the (i+1)th point.

[0021] In this embodiment, the byte number charNum where the bit is located is charNum=id. x / 8, the bit offset charBit is charBit=id x %8, the bit mask moveBit is moveBit=(1< <charBit)。

[0022] In this embodiment, the flag bit hasPt of the point is, hasPt=havePoint[charNum]&moveBit.

[0023] In this embodiment, as Figure 3 As shown, in step M3, the optimization of the sparsed radar point cloud using the improved transient search optimization algorithm includes: M31. Based on the data information of the radar point cloud after the current frame is sparsified, the population is initialized, the population parameters and the maximum number of iterations L are determined, and the data information of the initialized population is obtained. M32. Based on the data information of the initialized population, update the thermal resistance coefficient C of the population. l and T; M33. Update the population position W(X,C) l ,T); M34. Evaluate fitness and update optimal position; M35. Determine if the maximum number of iterations L is satisfied. If it is, output the optimal position and the global optimal solution, which are the final radar point cloud data information for the current frame. If it is not satisfied, return to step M32.

[0024] In this embodiment, the thermal resistivity C of the population l And T is, , , Where z is the attenuation coefficient variable changing from 2 to 0, k is a real number, and r1, r2 and r3 are random numbers in [0,1].

[0025] In this embodiment, the location of the population is W(X,C). l ,T) is, , Where X represents the individual parameters of the initialized population, and e represents the natural base.

[0026] Example 2: Based on the bit-based point cloud sparsification method of Example 1, the present invention will be further explained and described below.

[0027] like Figure 1 As shown, a bit-based method for sparsifying LiDAR point clouds is described, the method comprising: M1. Initialize the grid and acquire the LiDAR point cloud data information for the current frame; M2. Based on the initial grid, a bit-based LiDAR point cloud sparsification algorithm is used to traverse the data information of the LiDAR point cloud in the current frame, and the LiDAR point cloud is sparsified to obtain the data information of the LiDAR point cloud after sparsification in the current frame. M3. Based on the data information of the sparsed radar point cloud of the current frame, an improved transient search optimization algorithm is used to optimize the sparsed radar point cloud to obtain the final radar point cloud data information of the current frame.

[0028] 1) Initialize the raster Based on grid resolution (s x ,s y ,s z The raster range (x[minX, maxX], y[miny, maxY], z[minZ, maxZ]) is used to initialize the raster havePoint, with a quantity of (( maxX - minX ) / s x +1)* ((maxY- minY) / s y +1) *((maxZ-minZ) / s z +1).

[0029] 2) Bit-based point cloud sparsification (1) Calculate the raster coordinates (x, y) of the i-th point. Axis , y Axis , z Axis ) The coordinates of the i-th point: (pt) ix , pt iy , pt iz ) x Axis =(pt ix -minX) / s x y Axis =(pt iy -minY) / s y z Axis =(pt iz -minZ) / s z (2) Raster projection Calculate the raster index. id x =x Axis +y Axis *((maxX-minX) / s x +1)+z Axis *((maxX-minX) / s x +1) *((maxY-minY) / s y +1) Calculate the byte number of the bit (charNum), the bit offset (charBit), the bit mask (moveBit), and the flag indicating whether there is a point within the grid (hasPt). charNum=id x / 8 charBit=id x %8 moveBit=(1< <charBit) hasPt=havePoint[charNum] & moveBit; Determine whether the grid cell containing the current point has already been projected into a point cloud. If hasPt=0, it means that the current grid has not been projected into the point cloud, so the point is stored in the sparsed point cloud list (sparsedPoints); otherwise, if hasPt=1, the current grid already has a point, so the grid projection of the (i+1)th point is performed.

[0030] 3) Return the sparsed point cloud (1) Clear the projection flags in the grid havePoint. (2) Returns a list of sparsedpoints.

[0031] In this embodiment, the present invention provides a bit-based lidar point cloud sparsification system, including a computer device programmed or configured to perform the steps of the bit-based lidar point cloud sparsification method.

[0032] In this embodiment, the present invention provides a computer-readable storage medium storing a computer program programmed or configured to perform the bit-based lidar point cloud sparsification method described above.

[0033] Any references to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0034] In summary, this invention can not only significantly reduce the time and computing platform resource consumption of point cloud processing, but also greatly reduce the time and resource consumption of point cloud sparsification while maintaining the quality of point cloud.

[0035] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A bit-based method for sparsifying point clouds in lidar systems, characterized in that, The method includes: M1. Initialize the grid and acquire the LiDAR point cloud data information for the current frame; M2. Based on the initial grid, a bit-based LiDAR point cloud sparsification algorithm is used to traverse the data information of the LiDAR point cloud in the current frame, and the LiDAR point cloud is sparsified to obtain the data information of the LiDAR point cloud after sparsification in the current frame. M3. Based on the data information of the sparsed radar point cloud of the current frame, an improved transient search optimization algorithm is used to optimize the sparsed radar point cloud to obtain the final radar point cloud data information of the current frame.

2. The bit-based lidar point cloud sparsification method according to claim 1, characterized in that, In step M2, the step of using a bit-based lidar point cloud sparsification algorithm to traverse the lidar point cloud data information of the current frame and perform sparsification processing on the lidar point cloud includes: M21. Project the data information of the current frame's lidar point cloud onto the initialized grid, and calculate the i-th lidar point cloud (pt). ix , pt iy , pt iz The corresponding raster coordinates (x) Axis , y Axis , z Axis ), ， ， ， Where (minX,minY,minZ) are the lower boundary points of the initialized raster; M22. Determine whether the current grid contains points from the lidar point cloud based on bit method. If it does, return to step M21; otherwise, proceed to step M23. M23. Store the i-th point into the sparse point cloud queue to obtain the data information of the sparse radar point cloud of the current frame.

3. The bit-based lidar point cloud sparsification method according to claim 2, characterized in that, In step M22, the bit-based method of determining whether the current grid contains points from the lidar point cloud includes: M221. Based on the i-th lidar point cloud (pt) ix , pt iy , pt iz The corresponding raster coordinates (x) Axis , y Axis ,z Axis ), calculate the raster index id x , , Where (maxX,maxY,maxZ) are the upper boundary points of the initialized raster; M222. Calculate the byte number charNum where the bit is located, the bit offset charBit, the bit mask moveBit, and the flag bit hasPt indicating whether there is a point in the grid; M223. Determine whether the grid cell containing the current point has been projected into a point cloud. If hasPt = 0, it means that the current grid cell has not been projected into a point cloud. Then, store the point in the sparsed point cloud list. Otherwise, hasPt = 1, the current grid cell already contains a point, and perform grid projection on the (i+1)th point.

4. The bit-based lidar point cloud sparsification method according to claim 3, characterized in that: The byte number where the bit is located, charNum, is charNum=id. x / 8, the bit offset charBit is charBit=id x %8, the bit mask moveBit is moveBit=(1< <charBit)。 5. The bit-based lidar point cloud sparsification method according to claim 3, characterized in that: The flag bit hasPt of the point is, hasPt=havePoint[charNum]&moveBit.

6. The bit-based lidar point cloud sparsification method according to claim 1, characterized in that, In step M3, the optimization of the sparsed radar point cloud using the improved transient search optimization algorithm includes: M31. Based on the data information of the radar point cloud after the current frame is sparsified, the population is initialized, the population parameters and the maximum number of iterations L are determined, and the data information of the initialized population is obtained. M32. Based on the data information of the initialized population, update the thermal resistance coefficient C of the population. l and T; M33. Update the population position W(X,C) l ,T); M34. Evaluate fitness and update optimal position; M35. Determine if the maximum number of iterations L is satisfied. If it is, output the optimal position and the global optimal solution, which are the final radar point cloud data information of the current frame. If it is not satisfied, return to step M32.

7. The bit-based lidar point cloud sparsification method according to claim 6, characterized in that: The thermal resistance coefficient C of the population l And T is, ， ， Where z is the attenuation coefficient variable changing from 2 to 0, k is a real number, and r1, r2 and r3 are random numbers in [0,1].

8. The bit-based lidar point cloud sparsification method according to claim 7, characterized in that: The location of the population is W(X,C). l ,T) is, ， Where X represents the individual parameters of the initialized population, and e represents the natural base.

9. A bit-based lidar point cloud sparsification system, comprising computer equipment, characterized in that, The computer device is programmed or configured to perform the steps of the bit-based lidar point cloud sparsification method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is programmed or configured to perform the bit-based lidar point cloud sparsification method according to any one of claims 1 to 8.

Images