Method, apparatus, mobile device, and computer program product for dynamically adjusting a grid structure of a laser radar

Abstract

The application belongs to the technical field of laser radars, and provides a method for dynamically adjusting a grid structure of a laser radar and related equipment, which comprises the following steps: first, constructing a layered grid structure with a sensor of the laser radar as the center according to Kth frame point cloud data of the laser radar; traversing the Kth frame point cloud data to determine the density distribution of non-ground points in the Kth frame point cloud data; obtaining current (K+T)th frame point cloud data of the laser radar; if the difference between the two frames of point cloud data represents that the road surface range change of the driving environment where the mobile device carrying the laser radar is located reaches a preset road surface change threshold, then dynamically adjusting the aforementioned layered grid structure based on the density distribution of non-ground points in the Kth frame point cloud data to obtain a new layered grid structure; and further enabling a ground segmentation algorithm to adapt to the point cloud density distribution under different driving environments, and reducing the occurrence of the situation that the system incorrectly determines the upper plane of a low obstacle as ground point cloud.

Description

Technical Field

[0001] This application belongs to the field of lidar technology, and particularly relates to a method, apparatus, mobile device and computer program product for dynamically adjusting the grid structure of lidar. Background Technology

[0002] In outdoor environments, ground points constitute a large proportion of LiDAR point cloud data. Accurate segmentation of ground points is crucial for mobile device pose estimation, while also reducing the computational burden of processing non-ground points and improving segmentation accuracy. Existing ground segmentation methods mainly fall into three categories: 1) Grid-based methods, which simplify computation by projecting 3D point clouds onto 2D grids, but dimensionality reduction results in information loss, leading to undersegmentation in complex scenes; 2) Planar model fitting methods, which use mathematical models to fit the ground, but struggle to adapt to the undulations and obstacles of outdoor ground; 3) Region grid-based methods, which require no additional projection processing and are adapted to LiDAR scanning characteristics, but fixed grid sizes can affect computational efficiency when point cloud density is uneven. Therefore, a method for dynamically adjusting the LiDAR grid structure is urgently needed to balance segmentation accuracy and computational efficiency. Summary of the Invention

[0003] This application provides a method, apparatus, mobile device, and computer program product for dynamically adjusting the grid structure of a lidar, aiming to solve the technical problems of insufficient accuracy and low computational efficiency in existing outdoor point cloud ground segmentation technologies.

[0004] In a first aspect, embodiments of this application provide a method for dynamically adjusting the grid structure of a lidar, applied to a mobile device, comprising:

[0005] Based on the point cloud data of the Kth frame of the radar, a hierarchical grid structure is constructed with the position of the lidar sensor as the center. The hierarchical grid structure includes M concentric segments, and each concentric segment includes several grid cells. The lidar is deployed on the mobile device.

[0006] The point cloud data in each grid cell of the Kth frame point cloud data is traversed to determine the non-ground point density distribution in the Kth frame point cloud data.

[0007] Obtain the point cloud data of the lidar at time t in frame K+T, and obtain the difference between the point cloud data in frame K+T and the point cloud data in frame K.

[0008] If the difference represents a change in the road surface range of the driving environment in which the mobile device is located, reaching a preset road surface change threshold, then the layered grid structure is dynamically adjusted based on the non-ground point density distribution in the Kth frame point cloud data to obtain a new layered grid structure for the K+T frame point cloud data.

[0009] In a first possible implementation of the first aspect, the step of constructing a hierarchical mesh structure centered on the position of the lidar sensor, based on the Kth frame point cloud data of the lidar, includes...

[0010] Based on the Kth frame point cloud data of the lidar, M concentric segments are constructed with the position of the lidar sensor as the center and in ascending order of radius. Each concentric segment includes several layers of circular grids, and each layer of circular grids contains several grid cells. The radius represents the distance between the radial boundary of each concentric segment and the position of the sensor.

[0011] The dynamic adjustment of the layered grid structure based on the non-ground point density distribution in the point cloud data of frame K to obtain a new layered grid structure for the point cloud data of frame K+T includes:

[0012] Based on the non-ground point density distribution in the Kth frame point cloud data, the radial boundary of the innermost concentric segment in the layered grid structure is dynamically adjusted to obtain a new layered grid structure for the K+T frame point cloud data.

[0013] The innermost concentric segment is the concentric segment with the shortest radius among the M concentric segments.

[0014] In a first possible implementation of the first aspect, the step of dynamically adjusting the radial boundary of the innermost concentric segment of the layered grid structure based on the non-ground point density distribution in the Kth frame point cloud data to obtain a new layered grid structure for the K+Tth frame point cloud data includes:

[0015] Based on the density distribution of non-ground points in the point cloud data of the Kth frame, a density threshold is set;

[0016] Calculate the density of non-ground points in each grid cell of the point cloud data in frame K+1. If the density of non-ground points in the grid cells of a certain layer of the circular grid in the M concentric segments exceeds the density threshold, then set that layer of the circular grid as the new radial boundary of the innermost concentric segment.

[0017] The radial ranges of the M concentric segments are redivided according to the new boundary to obtain a new hierarchical grid structure for the point cloud data of the K+T frame;

[0018] The density of non-ground points in each grid cell is calculated using the following method:

[0019] Determine the total number of point clouds and the number of ground points within the grid cell;

[0020] The density of non-ground points in the grid cell is calculated based on the total number of point clouds and the number of ground points.

[0021] In a first possible implementation of the first aspect, after obtaining the new hierarchical mesh structure of the point cloud data in the K+T frame, the method further includes:

[0022] The grid cells of the m-th concentric segment in the point cloud data of the K+T-th frame are traversed to determine the polar coordinate index of the current grid cell; and the current radius of the m-th concentric segment is obtained.

[0023] Find the historical downsampling parameters of the historical grid cell corresponding to the polar coordinate index in the point cloud data of the Kth frame;

[0024] Based on the current radius and the feedback of the historical downsampling parameters, set the current downsampling ratio of the current grid cell;

[0025] The point cloud density of the current grid cell is reduced by the current downsampling ratio.

[0026] In a first possible implementation of the first aspect, the method further includes:

[0027] Count the number of grid cells contained in the m-th concentric segment;

[0028] The step of finding the historical downsampling parameter feedback of the historical grid cell corresponding to the polar coordinate index in the point cloud data of the Kth frame includes:

[0029] Find the historical radial distance of the outer boundary of the historical grid cell corresponding to the polar coordinate index in the Kth frame point cloud data relative to the sensor of the lidar;

[0030] The step of setting the current downsampling ratio of the current grid cell based on the current radius and the historical downsampling parameters includes:

[0031] Based on the number of grid cells contained in the m-th concentric segment, the current radius, and the historical radial distance, calculate the outer boundary of the current grid cell of the m-th concentric segment of the K+T frame point cloud data relative to the current radial distance of the lidar sensor.

[0032] The current downsampling ratio of the current grid cell is set according to the current radial distance.

[0033] In a first possible implementation of the first aspect, the current radial distance is calculated using the following formula:

[0034]

[0035] Among them, l preΔL is the historical radial distance ΔL of the outer boundary of the historical grid cell corresponding to the polar coordinate index in the Kth frame of point cloud data relative to the sensor of the lidar. m For the m-th concentric segment Z m The current radius, N r,m For the m-th concentric segment Z m The number of grid cells contained; l represents the outer boundary of the current grid cell of the m-th concentric segment of the point cloud data in frame K+T relative to the current radial distance of the lidar sensor.

[0036] In a first possible implementation of the first aspect, the current downsampling ratio is calculated using the following formula:

[0037]

[0038] P c The current downsampling ratio of the current grid cell is represented by , e represents the natural constant, and h represents the attenuation adjustment factor.

[0039] The beneficial effects of the first aspect of this application are as follows: First, based on the Kth frame point cloud data of the LiDAR, a hierarchical grid structure centered on the sensor position of the LiDAR is constructed; the point cloud data in each grid cell of the Kth frame point cloud data is traversed to determine the non-ground point density distribution in the Kth frame point cloud data; the K+T frame point cloud data of the LiDAR at time t is obtained, and the difference between the K+T frame point cloud data and the Kth frame point cloud data is obtained. If the difference indicates that the change in the road surface range of the driving environment of the mobile device equipped with the LiDAR reaches a preset road surface change threshold, then based on the non-ground point density distribution in the Kth frame point cloud data, the aforementioned hierarchical grid structure is dynamically adjusted to obtain a new hierarchical grid structure for the K+T frame point cloud data; thereby enabling the ground segmentation algorithm to adapt to the point cloud density distribution under different driving environments, and reducing the probability of the system mistakenly identifying the upper surface of low obstacles as ground point clouds.

[0040] Secondly, this application also proposes a device for dynamically adjusting the grid structure of a lidar, comprising:

[0041] Initialization module: used to construct a hierarchical grid structure with the position of the lidar sensor as the center based on the point cloud data of the Kth frame of the lidar. The hierarchical grid structure includes M concentric segments, and each concentric segment includes several grid cells; wherein the lidar is deployed on the mobile device.

[0042] Feedforward processing module: used to traverse the point cloud data in each grid cell of the Kth frame point cloud data to determine the non-ground point density distribution in the Kth frame point cloud data;

[0043] Environmental monitoring module: used to acquire the point cloud data of the lidar at time t in the K+T frame, and to acquire the difference between the point cloud data in the K+T frame and the point cloud data in the K frame;

[0044] Dynamic adjustment module: If the difference characterizes the change in the road surface range of the driving environment where the mobile device is located, reaching a preset road surface change threshold, then based on the non-ground point density distribution in the Kth frame point cloud data, the layered grid structure is dynamically adjusted to obtain a new layered grid structure for the K+T frame point cloud data.

[0045] Thirdly, this application also proposes a mobile device, the mobile device including a lidar interface, a memory, a processor, and an executable program for dynamically adjusting the mesh structure of a lidar stored in the memory and executable on the processor; when the processor executes the executable program for dynamically adjusting the mesh structure of the lidar, it implements the steps of the method for dynamically adjusting the mesh structure of the lidar as described in the first aspect above.

[0046] Fourthly, this application also proposes a computer program product that stores an execution program for dynamically adjusting the grid structure of a lidar, wherein when the execution program for dynamically adjusting the grid structure of the lidar is executed by a processor, it implements the steps of the method for dynamically adjusting the grid structure of the lidar as described in the first aspect above.

[0047] It is understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the first or second aspects mentioned above, and will not be repeated here. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a schematic diagram of the structure of the mobile device provided in the embodiments of this application;

[0050] Figure 2 This is a flowchart illustrating a method for dynamically adjusting the grid structure of a lidar according to an embodiment of this application.

[0051] Figure 3 This is a schematic diagram of a layered grid structure with the position of the lidar sensor as the center, provided in an embodiment of this application.

[0052] Figure 4This is a schematic diagram illustrating environmental changes as the vehicle travels in an outdoor setting, as provided in this application.

[0053] Figure 5 This is a schematic flowchart of another embodiment of a method for dynamically adjusting the grid structure of a lidar provided in this application;

[0054] Figure 6 This is a schematic diagram of a device for dynamically adjusting the grid structure of a lidar according to an embodiment of this application. Detailed Implementation

[0055] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0056] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0057] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0058] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0059] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0060] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0061] The applicant found that in outdoor environments, ground points constitute a large proportion of the point cloud acquired by lidar. Accurate ground point segmentation and extraction are of great significance. Ground points have distinct features and, as a plane, can provide good constraints for pose estimation of mobile devices. Furthermore, after extracting ground points, the clustering and segmentation of non-ground points require less computation and have higher segmentation accuracy.

[0062] Existing ground segmentation methods typically fall into three categories:

[0063] The first approach is a grid-based method, which is relatively simple to compute. It involves projecting a 3D point cloud into a 2D space and then performing the calculation. Because the computational load is small, it can meet the real-time requirements. However, dimensionality reduction will lose some information. Therefore, this type of method will have the problem of undersegmentation in highly complex scenarios, that is, the ground segmentation is incomplete.

[0064] The second method is based on a planar model fitting method. This method fits the ground to a simple mathematical model and finds the planar model that is most similar to the ground. This allows the points on the plane that are close to the surrounding area to be segmented. However, the outdoor ground is often not a plane. There will be undulations, slopes, or obstacles.

[0065] The third approach is a region-grid-based method. This method has good performance, and the grid division can naturally adapt to the scanning characteristics of LiDAR. It does not require additional processing of the point cloud, such as projection. However, when the density of the point cloud is different, using a fixed-size grid will significantly affect the allocation of computing resources.

[0066] To address the aforementioned technical problems of insufficient accuracy and low computational efficiency in existing outdoor point cloud ground segmentation technologies, this application provides a method, apparatus, mobile device, and computer program product for dynamically adjusting the grid structure of a lidar system.

[0067] refer to Figure 1 , Figure 1This is a schematic diagram of the structure of a mobile device according to an embodiment of this application. The mobile device includes: at least one processor 1001, a communication bus 1002, a memory 1005, and an executable program 1006 for dynamically adjusting the grid structure of a lidar, stored in the memory 1005 and executable on the at least one processor 1001. When the processor 1001 executes the executable program 1006 for dynamically adjusting the grid structure of the lidar, it implements the steps in the method embodiment of dynamically adjusting the grid structure of the lidar according to this application. Figure 1 The mobile devices shown can be mobile vehicles, such as cars and motorcycles, or intelligent mobile robotic devices such as robot vacuum cleaners.

[0068] The mobile device also includes a lidar interface 1003 for connecting to a lidar device, through which the mobile device can transmit relevant data to the lidar device.

[0069] The mobile device may also include a camera interface for acquiring road surface images of the driving environment; in some embodiments, the camera interface may be connected to the internal image sensor of the mobile device, and in other embodiments, the camera interface may be connected to an external camera device of the mobile device.

[0070] Those skilled in the art will understand that Figure 1 This is merely an example of a mobile device and does not constitute a limitation on mobile devices. It may include more or fewer components than shown, or combine certain components, or use different components.

[0071] The processor 1001 may be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0072] In some embodiments, the memory 1005 may be an internal storage unit of the mobile device, such as a hard drive or memory. In other embodiments, the memory 1005 may be an external storage device of the mobile device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory 1005 may include both internal and external storage units of the mobile device. The memory 1005 is used to store the operating system and the computer-executable program for the method of dynamically adjusting the grid structure of the LiDAR.

[0073] The mobile device described in this application embodiment is exemplified by a vehicle. When the processor 1001 executes the computer program 1006, it implements the method steps of dynamically adjusting the grid structure of the lidar, as described in this application. Figure 2 As shown, steps A1 to A4 are included:

[0074] A1 (Initialization Phase): Based on the Kth frame point cloud data of the lidar, construct a hierarchical grid structure with the position of the lidar sensor as the center. The hierarchical grid structure includes M concentric segments, and each concentric segment includes several grid units.

[0075] Understandably, when a mobile device is in motion, the lidar it carries begins to work and can detect the surrounding point cloud in real time, transmitting the detected point cloud data to the mobile device's processor.

[0076] In one embodiment, based on the Kth frame point cloud data of the lidar, M concentric segments (Z1, Z2, Z3…Z4) are constructed with the lidar sensor position as the center and in ascending order of radius. m-1 Z m Each concentric segment includes several layers of circular grids, each layer of circular grids contains several grid cells, and each grid cell can store a certain amount of point cloud; the radius represents the distance between the radial boundary of each concentric segment and the position of the sensor;

[0077] For example, the Kth frame is the point cloud data from a previous frame, such as... Figure 3 As shown, with the sensor position O of the lidar as the center, M=4 concentric segments are constructed in ascending order of radius. These are: the first concentric segment Z1 ( Figure 2 (green area), second concentric section Z2 ( Figure 2 (red area), third concentric segment Z3 ( Figure 2 (blue area), fourth concentric segment Z4 ( Figure 2 The gray area); specifically, firstly, the corresponding fourth concentric segment Z4 can be set according to the maximum measurement distance of the radar sensor. The radius of the fourth concentric segment Z4 is the maximum range of the layered grid structure, for example, the radius of Z4 is 20 meters; the corresponding first concentric segment Z1 can be set according to the minimum measurement distance of the radar sensor. The radius of the fourth concentric segment Z1 (i.e., radius) is the minimum value of the layered grid structure, for example, the radius of Z1 is 3 meters; and the second concentric segment Z2 and the third concentric segment Z3 can be set according to the size of the first concentric segment Z1 and the fourth concentric segment Z4 to equally divide the area corresponding to the middle radial length.

[0078] Understandably, the larger the radius of a concentric segment (i.e., the farther its outer boundary is from the sensor), the sparser the point cloud collected over the same radial span. Correspondingly, the greater the radial distance between the grid cells in each concentric segment and the sensor, the sparser the collected point cloud becomes. Therefore, within the concentric segment, the density of radial grid cells is further layered radially, meaning that grid cells in layers farther from the sensor are larger. This setup achieves better ground estimation, thereby improving overall segmentation accuracy.

[0079] A2 (Feedforward Processing Stage): Traverse the point cloud data in each grid cell of the Kth frame point cloud data to determine the non-ground point density distribution in the Kth frame point cloud data;

[0080] A3 (Environmental Monitoring Phase): Acquire the point cloud data of the lidar at time t in frame K+T, and obtain the difference between the point cloud data of frame K+T and the point cloud data of frame K.

[0081] Understandably, this application embodiment monitors the point cloud data of the LiDAR in real time at various moments. If, at a certain moment t, a difference is detected between the corresponding point cloud data of the K+T frame and the previous K frame point cloud data (the K frame is a frame preceding the K+T frame), and the difference indicates that the change in the road surface range of the driving environment where the mobile device is located has reached a preset road surface change threshold, then... Figure 4 As shown, step A4 needs to be performed at this moment to adjust the current hierarchical mesh structure of the lidar:

[0082] A4 (Dynamic Adjustment Stage): Based on the non-ground point density distribution in the Kth frame point cloud data, the hierarchical grid structure is dynamically adjusted to obtain a new hierarchical grid structure for the K+T frame point cloud data.

[0083] Understandably, in the outdoor application scenario of this application embodiment, when the vehicle is traveling in an area with significant changes in road surface area at time t, directly using the fixed grid size and downsampling parameters of the layered grid structure constructed in the previous k-th frame may affect the accuracy of ground point segmentation, potentially impacting the vehicle's mapping and positioning accuracy at time t and ultimately affecting the stability of vehicle navigation. In this case, by executing step A4 to dynamically adjust the layered grid structure formed in the previous k-th frame, a new layered grid structure is created that is more suitable for the environment at time t, making the vehicle's autonomous navigation more stable.

[0084] In one embodiment, the radial boundary of the innermost concentric segment Z1 in the layered grid structure of step A1 can be dynamically adjusted based on the non-ground point density distribution in the point cloud data of frame K to obtain a new layered grid structure for the point cloud data of frame K+T. It is understood that the innermost concentric segment Z1 is the concentric segment with the shortest radius among the aforementioned four (Z1, Z2, Z3, Z4) concentric segments. Adjusting only the innermost first concentric segment Z1 will adaptively adjust the second concentric segment Z2, the third concentric segment Z3, and the fourth concentric segment Z4.

[0085] For example, this application can set a density threshold T based on the non-ground point density distribution in the point cloud data of the Kth frame (previous frame); calculate the density of non-ground points in each grid cell of the point cloud data of the K+Tth frame (current frame); if the density of non-ground points in a certain grid cell of a certain layer of the aforementioned M concentric segments (Z1, Z2, Z3, Z4) exceeds the set density threshold T for the first time, then set that layer of the concentric grid as the new radial boundary of the new innermost concentric segment Z1'; then redivide the radial range of the m concentric segments according to the new boundary to obtain a new layered grid structure of the point cloud data of the K+Tth frame (current frame). For example, in the point cloud data of the (current frame) K+T frame, if the non-ground point density of a certain grid cell in a certain layer of the circular grid in the second concentric segment Z2 of the aforementioned m concentric segments (Z1, Z2, Z3, Z4) exceeds the set density threshold T for the first time, then this layer of circular grid is set as the new radial boundary of the new innermost concentric segment Z1'; then the radial range of the four concentric segments is re-divided based on the new boundary to obtain the new layered grid structure (Z1', Z2', Z3', Z4') of the point cloud data of the (current frame) K+T frame;

[0086] For example, in this application embodiment, the total number of point clouds and the number of ground points in the grid cell can be determined; then, based on the total number of point clouds and the number of ground points, the density of non-ground points in the grid cell can be calculated.

[0087] In practical implementation, the density of non-ground points in each grid cell can be calculated using the following formula 1:

[0088]

[0089] In this embodiment, a polar coordinate index bin is assigned to each grid cell in each layer of the annular grid. r,θ ; r represents the radial distance from the acquisition point corresponding to the grid cell to the origin O of the lidar sensor coordinate system, and θ is the polar angle in the polar coordinate system. n r,θ N represents the number of ground points. r,θ ρ represents the total number of point clouds within a single grid cell. r,θ This represents the density of non-ground points in the grid cell.

[0090] In this embodiment, during the process of traversing the point cloud data in each grid cell, the total number N of point clouds in a single traversed grid cell can first be determined. r,θ Then determine the number n of ground points within a single grid cell traversed. r,θ Based on the total number N of point clouds within a single grid cell r,θ and the number of ground points n within the grid cell r,θ The density ρ of non-ground points in the grid cell is calculated using Formula 1. r,θ .

[0091] Understandably, in applications where mobile devices travel on complex road surfaces, over 90% of ground points can be distributed within 20 meters of the mobile device, even in situations where the road is wide or narrow. When there are numerous low obstacles on both sides of the road, the density distribution of ground points will vary. Existing ground point segmentation technologies sometimes incorrectly identify the upper plane of low obstacles as ground point clouds. To reduce this, by enabling the LiDAR to better adapt to the layered mesh structure model, this embodiment performs steps A1 to A4 above, adjusting the radius of the innermost concentric segment Z1 using an adaptive adjustment method. This reduces the likelihood of incorrectly identifying the upper plane of low obstacles as ground point clouds. When significant changes in the road surface area are detected in the current driving environment, the boundary of the layered mesh structure of the LiDAR's current frame point cloud data can be dynamically adjusted based on the processing results of the previous frame point cloud data to adapt to the point cloud density distribution under different driving environments.

[0092] Furthermore, in some embodiments, after obtaining the new hierarchical grid structure of the point cloud data of the K+T frame in step A4, principal component analysis (PCA) can be used to reduce the dimensionality of the point cloud and extract height information within each grid cell (bin). Then, after the principal component analysis, the innermost concentric segment Z1 may have insufficient spatial representation due to its small size, so it is necessary to increase the size of each grid cell of the innermost concentric segment Z1. However, after the size is increased, the point cloud density of the innermost concentric segment Z1 increases significantly. If the plane is directly fitted, the computational load will surge due to outliers (such as vehicles, pedestrians, and other irregular obstacles).

[0093] To address the aforementioned technical issues, before performing principal component analysis on each grid cell to reduce the dimensionality of the point cloud, refer to... Figure 5 In this embodiment of the application, the following steps A51 to A54 can be performed to perform percentage downsampling on the grid cells of each concentric segment, which effectively reduces computational complexity and improves fitting efficiency while retaining key ground points:

[0094] A51: Traverse the grid cells of the m-th concentric segment in the point cloud data of the K+T-th frame, and determine the polar coordinate index bin of the current grid cell bin that is traversed. r,θ And obtain the current radius ΔL of the m-th concentric segment. m ;

[0095] For example, in this embodiment, each grid cell of the m-th concentric segment in the K+T-th frame (the current frame) is traversed to determine the current radius ΔL of the m-th concentric segment. m Simultaneously, the number N of grid cells contained in the m-th concentric segment can also be counted. r,m ;

[0096] A52: Find the point cloud data in frame K that matches the polar coordinate index bin. r,θ Feedback of historical downsampling parameters for the corresponding historical grid cells;

[0097] Understandably, regardless of how the size of the layered grid structure changes, and regardless of how the radius of each concentric segment changes, the position of the polar coordinate index of a single grid cell remains unchanged. Therefore, it is possible to determine the position of the polar coordinate index bin of a grid cell in frame K+T (the current frame). r,θ To find the polar coordinate index bin r,θ The historical radial distance l during the last downsampling in the Kth frame (the previous frame) pre ;

[0098] A53: Based on the current radius ΔL mBased on the feedback of the historical downsampling parameters, the current downsampling ratio of the current grid cell bin is set;

[0099] Among them, the closer the grid cell is to the sensor, the higher the downsampling ratio;

[0100] In one embodiment, the number N of grid cells contained in the m-th concentric segment can be used as a basis. r,m Current radius ΔL m Historical radial distance l pre Calculate the current radial distance l of the outer boundary of the current grid cell bin of the m-th concentric segment of the point cloud data in frame K+T relative to the lidar sensor; for example, the current radial distance l can be calculated using the following formula 2:

[0101]

[0102] Then, set the current downsampling ratio of the current grid cell bin according to the current radial distance l; for example, the current downsampling ratio P can be calculated using the following formula 3. c :

[0103]

[0104] In this embodiment, e represents the natural constant (≈2.71828), P c This represents the current downsampling ratio, i.e., the ratio of discarded points in the current grid cell bin downsampling. Its range can be set to 0 ≤ P. c ≤0.5; e -l / h The exponential decay model represents the attenuation characteristics of the downsampling ratio as the distance increases; h represents the attenuation adjustment factor, and relevant personnel can set the value of h in advance according to the performance parameters of the lidar.

[0105] It is understandable that the exponential decay model in this embodiment expresses the density distribution law of the lidar point cloud (dense near and sparse far) through an exponential function; exponential decay can quickly reduce the downsampling intensity at nearby locations and avoid excessive sparseness of the point cloud at distant locations.

[0106] For example, for the exponential function e -l / h Through certain experiments, this application embodiment shows that setting h to 10 can effectively control the decay rate, i.e., the exponential function is e -l / 10 For example, a larger denominator (e.g., 20) results in slower attenuation and a smoother change in the downsampling ratio; conversely, a smaller denominator (e.g., 5) leads to faster attenuation and more aggressive downsampling at close range. Therefore, setting the denominator to 10 in this embodiment effectively controls the attenuation rate, balances downsampling requirements at both near and far ranges, and is adaptable to the characteristics of different LiDAR sensors. For the exponential function e^(-1 / 2)... -l / 10 ,Right now This can be understood as limiting the downsampling rate to a reasonable range (maximum 50%) to avoid excessive deletion of valid points; it can also be understood that, since the value range is (0,1], the maximum value of 1 is taken when l=0 (i.e., the closest point to the sensor). As l increases Attenuation, gradually decreasing (e.g., when l = 10 meters, Therefore, taking half of the exponential function can limit the downsampling ratio to a reasonable range, with the maximum discard rate limited to 50%, avoiding excessive deletion of valid points, which is more in line with actual needs (retaining at least 50% of the points to ensure fitting accuracy). This achieves the effect that the point cloud density is high in the near area, requiring a higher downsampling ratio to reduce the amount of computation, while the point cloud is sparse in the far area, requiring a lower downsampling ratio to retain key features.

[0107] A54: Reduce the point cloud density of the current grid cell bin by the current downsampling rate.

[0108] Before performing principal component analysis on each grid cell to reduce the dimensionality of the point cloud, the current grid cell of each concentric segment being traversed is downsampled by a percentage according to the current downsampling ratio. This effectively reduces computational complexity and improves fitting efficiency while preserving key ground points.

[0109] The mobile device described in this application uses a vehicle as an example. For instance, when the vehicle is in an outdoor scenario and the road surface area of ​​the vehicle's driving environment changes significantly, the method steps for dynamically adjusting the grid structure of the LiDAR provided in this application can be executed.

[0110] When navigating to areas with significant road surface variations, directly applying a fixed grid size and downsampling parameters may affect the accuracy of ground point segmentation, which in turn affects the accuracy of mapping and positioning, ultimately impacting the stability of vehicle navigation. In such cases, the steps described below can achieve more environmentally adaptable and stable autonomous vehicle navigation.

[0111] In one embodiment, such as Figure 6 As shown, the present invention also provides a device for dynamically adjusting the grid structure of a lidar, comprising:

[0112] Initialization module 10: is used to construct a hierarchical grid structure with the position of the lidar sensor as the center based on the point cloud data of the Kth frame of the lidar. The hierarchical grid structure includes M concentric segments, and each concentric segment includes several grid cells; wherein the lidar is deployed on the mobile device.

[0113] Feedforward processing module 20: used to traverse the point cloud data in each grid cell of the Kth frame point cloud data to determine the non-ground point density distribution in the Kth frame point cloud data;

[0114] Environmental monitoring module 30: used to acquire the point cloud data of the lidar at time t in frame K+T, and to acquire the difference between the point cloud data in frame K+T and the point cloud data in frame K;

[0115] Dynamic adjustment module 40: If the difference characterizes the change in the road surface range of the driving environment where the mobile device is located, and reaches a preset road surface change threshold, then based on the density distribution of non-ground points in the point cloud data of the Kth frame, the layered grid structure is dynamically adjusted to obtain a new layered grid structure for the point cloud data of the K+Tth frame.

[0116] It should be noted that the above-mentioned device can be understood as a chip in the mobile device; the information interaction, execution process and other contents between the above-mentioned devices / units are based on the same concept as the method embodiment of this application, and their specific functions and technical effects can be found in the method embodiment section, which will not be repeated here.

[0117] Furthermore, this application also provides a computer program product that stores an execution program for dynamically adjusting the grid structure of a lidar. When the execution program for dynamically adjusting the grid structure of the lidar is executed by a processor, it implements the steps of the method embodiment for dynamically adjusting the grid structure of the lidar as described above.

[0118] Furthermore, it should be noted that if the integrated unit described in this application is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0119] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0120] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0121] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0122] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0123] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for dynamically adjusting the grid structure of a lidar, applied to mobile devices, characterized in that, include: Based on the Kth frame point cloud data of the lidar, a hierarchical grid structure is constructed with the position of the lidar sensor as the center. The hierarchical grid structure includes M concentric segments, and each concentric segment includes several grid cells. The lidar is deployed on the mobile device. The point cloud data in each grid cell of the Kth frame point cloud data is traversed to determine the non-ground point density distribution in the Kth frame point cloud data. Obtain the point cloud data of the lidar at time t in frame K+T, and obtain the difference between the point cloud data in frame K+T and the point cloud data in frame K; where K, t, and T are all positive integers. If the difference represents a change in the road surface range of the driving environment in which the mobile device is located, reaching a preset road surface change threshold, then the layered grid structure is dynamically adjusted based on the non-ground point density distribution in the Kth frame point cloud data to obtain a new layered grid structure for the K+T frame point cloud data.

2. The method as described in claim 1, characterized in that, The step of constructing a hierarchical mesh structure centered on the sensor position of the lidar based on the Kth frame point cloud data of the lidar includes constructing M concentric segments with the sensor position of the lidar as the center, arranged in ascending order of radius, based on the Kth frame point cloud data of the lidar. Each concentric segment includes several layers of annular meshes, and each annular mesh contains several mesh cells. The radius represents the distance between the radial boundary of each concentric segment and the sensor position. The dynamic adjustment of the layered grid structure based on the non-ground point density distribution in the point cloud data of frame K to obtain a new layered grid structure for the point cloud data of frame K+T includes: Based on the non-ground point density distribution in the Kth frame point cloud data, the radial boundary of the innermost concentric segment in the layered grid structure is dynamically adjusted to obtain a new layered grid structure for the K+T frame point cloud data. The innermost concentric segment is the concentric segment with the shortest radius among the M concentric segments.

3. The method as described in claim 2, characterized in that, The step of dynamically adjusting the radial boundary of the innermost concentric segment of the layered grid structure based on the non-ground point density distribution in the Kth frame point cloud data to obtain a new layered grid structure for the K+Tth frame point cloud data includes: Based on the density distribution of non-ground points in the point cloud data of the Kth frame, a density threshold is set; Calculate the density of non-ground points in each grid cell of the point cloud data in frame K+1. If the density of non-ground points in the grid cells of a certain layer of the circular grid in the M concentric segments exceeds the density threshold, then set that layer of the circular grid as the new radial boundary of the innermost concentric segment. The radial ranges of the M concentric segments are redivided according to the new radial boundary to obtain a new hierarchical grid structure for the point cloud data of the K+T frame; The density of non-ground points in each grid cell is calculated using the following method: Determine the total number of point clouds and the number of ground points within the grid cell; The density of non-ground points in the grid cell is calculated based on the total number of point clouds and the number of ground points.

4. The method according to any one of claims 1 to 3, characterized in that, After obtaining the new hierarchical mesh structure of the point cloud data in the K+T frame, the method further includes: The grid cells of the m-th concentric segment in the point cloud data of the K+T-th frame are traversed to determine the polar coordinate index of the current grid cell; and the current radius of the m-th concentric segment is obtained. Find the historical downsampling parameters of the historical grid cell corresponding to the polar coordinate index in the point cloud data of the Kth frame; Based on the current radius and the feedback of the historical downsampling parameters, set the current downsampling ratio of the current grid cell; The point cloud density of the current grid cell is reduced by the current downsampling ratio.

5. The method as described in claim 4, characterized in that, The method further includes: Count the number of grid cells contained in the m-th concentric segment; The step of finding the historical downsampling parameter feedback of the historical grid cell corresponding to the polar coordinate index in the point cloud data of the Kth frame includes: Find the historical radial distance of the outer boundary of the historical grid cell corresponding to the polar coordinate index in the Kth frame point cloud data relative to the sensor of the lidar; The step of setting the current downsampling ratio of the current grid cell based on the current radius and the historical downsampling parameters includes: Based on the number of grid cells contained in the m-th concentric segment, the current radius, and the historical radial distance, calculate the outer boundary of the current grid cell of the m-th concentric segment of the K+T frame point cloud data relative to the current radial distance of the lidar sensor. Set the current downsampling ratio of the current grid cell according to the current radial distance.

6. The method as described in claim 5, characterized in that, The current radial distance is calculated using the following formula: Among them, l pre ΔL is the historical radial distance ΔL of the outer boundary of the historical grid cell corresponding to the polar coordinate index in the Kth frame point cloud data relative to the sensor of the lidar. m For the m-th concentric segment Z m The current radius, N r,m For the m-th concentric segment Z m The number of grid cells included; l represents the outer boundary of the current grid cell of the m-th concentric segment of the point cloud data in frame K+T relative to the current radial distance of the lidar sensor.

7. The method as described in claim 6, characterized in that, The current downsampling ratio is calculated using the following formula: P c The current downsampling ratio of the current grid cell is represented by , e represents the natural constant, and h represents the attenuation adjustment factor.

8. A device for dynamically adjusting the grid structure of a lidar, characterized in that, include: Initialization module: used to construct a hierarchical grid structure with the position of the lidar sensor as the center based on the point cloud data of the Kth frame of the lidar. The hierarchical grid structure includes M concentric segments, and each concentric segment includes several grid cells; wherein the lidar is deployed on a mobile device. Feedforward processing module: used to traverse the point cloud data in each grid cell of the Kth frame point cloud data to determine the non-ground point density distribution in the Kth frame point cloud data; Environmental monitoring module: used to acquire the point cloud data of the lidar at time t in the K+T frame, and to acquire the difference between the point cloud data in the K+T frame and the point cloud data in the K frame; Dynamic adjustment module: If the difference characterizes the change in the road surface range of the driving environment where the mobile device is located, reaching a preset road surface change threshold, then based on the non-ground point density distribution in the Kth frame point cloud data, the layered grid structure is dynamically adjusted to obtain a new layered grid structure for the K+T frame point cloud data.

9. A mobile device, characterized in that, The mobile device includes a lidar interface, a memory, a processor, and an executable program for dynamically adjusting the lidar mesh structure stored in the memory and executable on the processor; when the processor executes the executable program for dynamically adjusting the lidar mesh structure, it implements the steps of the method for dynamically adjusting the lidar mesh structure as described in any one of claims 1-7.

10. A computer program product, said computer program product storing an executable program for dynamically adjusting the grid structure of a lidar, characterized in that, When the execution program for dynamically adjusting the grid structure of the lidar is executed by the processor, it implements the steps of the method for dynamically adjusting the grid structure of the lidar as described in any one of claims 1-7.

Images