Laser radar ground segmentation algorithm, electronic equipment and storage medium

Through the two-dimensional projection and elevation pre-classification of lidar point clouds, combined with conditional random field model and graph cutting algorithm, the efficiency and accuracy of lidar ground segmentation in complex terrain is solved, and more efficient and accurate ground segmentation is achieved , improving the adaptability and performance of the autonomous driving system.

CN119048536APending Publication Date: 2024-11-29CHANGSHU INSTITUTE OF TECHNOLOGY

Patent Information

Application Number
CN202411178708.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The prior art is difficult to achieve efficient and accurate lidar ground segmentation under complex terrain conditions, resulting in the impact of subsequent perception and decision-making tasks and requires a large number of parameter debugging.

Method used

The two-dimensional projection and elevation pre-classification of lidar point clouds are used, combined with the conditional random field model and graph cutting algorithm to perform ground segmentation, which improves segmentation accuracy and real-timeness, and reduces the calculation amount and parameter adjustment requirements.

Benefits of technology

Provide more reliable perception, positioning and navigation support in complex environments, improving the adaptability and performance of autonomous driving systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048536A_ABST
    Figure CN119048536A_ABST
Patent Text Reader

Abstract

The invention provides a laser radar ground segmentation algorithm, electronic equipment and a storage medium. The segmentation algorithm comprises the steps that point cloud information of a target area is acquired through a laser radar, a two-dimensional laser point cloud depth image is obtained through laser radar spherical projection, and each pixel of the two-dimensional laser point cloud depth image corresponds to one point cloud; segmenting the point cloud information into a non-ground point, an unknown classification point and a ground point according to the two-dimensional laser point cloud depth image; marking the non-ground point, the unknown classification point and the ground point in the two-dimensional laser point cloud depth image; constructing a graph model for the marked two-dimensional laser point cloud depth image based on a conditional random field; and carrying out ground segmentation on the graph model through a graph segmentation algorithm.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to a laser radar ground segmentation algorithm, electronic equipment and storage medium. Background Art

[0002] As one of the preprocessing parts of many point cloud perception tasks, ground segmentation is a key link in autonomous driving technology. It can give ground attributes to the 3D lidar point cloud of autonomous driving vehicles, laying the foundation for tasks such as robot autonomous driving and collision-free navigation. In these tasks, intelligent vehicles need to rely on accurate segmentation technology to identify various meaningful objects in the environment, and then track, ignore or bypass these objects in subsequent tasks, or use them in the positioning and mapping process. In order to achieve fully autonomous navigation tasks, intelligent vehicles are usually equipped with a variety of sensors for perception. For example, cameras can provide high-resolution image information, lidar can accurately measure distance, and 4D millimeter-wave radar provides richer environmental perception data. These sensors have their own advantages and disadvantages, and their combined use can improve the accuracy and robustness of environmental perception.

[0003] Among them, 3D LiDAR has become an indispensable part of the perception task of autonomous vehicles due to its unique advantages. 3D LiDAR is a relatively independent sensor that can provide accurate distance information. Its ranging accuracy is not affected by changes in ambient light, and it also has a wide field of view of up to 360 degrees. These characteristics make LiDAR important in the perception task of autonomous driving. However, these advantages also mean that the amount of data generated by LiDAR is very large. These point cloud data not only contain information about various objects in the surrounding environment, but also ground information. Before performing subsequent high-precision object segmentation, accurate object recognition, and robust tracking tasks, these point cloud data must first be ground segmented to remove redundant ground point clouds.

[0004] In order to improve the efficiency and accuracy of ground segmentation, researchers have proposed a variety of algorithms and techniques. Elevation-based methods usually distinguish between ground points and non-ground points by setting an elevation threshold, but in terrain with obvious slope changes, this method is prone to misidentifying some non-ground points as ground points, or vice versa. Plane fitting-based methods attempt to identify ground points by fitting a plane model, but in complex terrain, the accuracy of plane fitting will also drop significantly.

[0005] These problems not only affect the accuracy of the segmentation results, but also affect subsequent perception and decision-making tasks due to too many error points. Therefore, developing a ground segmentation algorithm that can maintain high efficiency and accuracy under various complex terrain conditions without requiring a large number of parameter adjustments is one of the important challenges facing the current ground segmentation task.

[0006] In order to solve the above problems, a laser radar ground segmentation algorithm, an electronic device and a storage medium are proposed here. Summary of the invention

[0007] The embodiments of the present disclosure aim to solve at least one of the technical problems existing in the prior art, and provide a laser radar ground segmentation algorithm, electronic device and storage medium, which effectively improves the accuracy and real-time performance of ground segmentation through two-dimensional projection and elevation pre-classification of laser point clouds, combined with conditional random field model and graph cut algorithm, reduces the amount of calculation and parameter adjustment requirements, and thus provides more reliable perception, positioning and navigation support in different environments. This innovative method overcomes the limitations of traditional methods under the framework of combining preliminary rough segmentation with fine segmentation, and improves the adaptability and performance of autonomous driving systems in complex environments.

[0008] In a first aspect of the present disclosure, a laser radar ground segmentation algorithm is provided, wherein the segmentation algorithm comprises: The point cloud information of the target area is obtained by laser radar, and a two-dimensional laser point cloud depth image is obtained by laser radar spherical projection, wherein each pixel of the two-dimensional laser point cloud depth image corresponds to a point cloud; Segmenting the point cloud information into non-ground points, unknown classification points, and ground points according to the two-dimensional laser point cloud depth image; Marking the non-ground points, unknown classification points, and ground points in the two-dimensional laser point cloud depth image; Constructing a graph model based on the labeled two-dimensional laser point cloud depth image based on the conditional random field; The graph model is ground segmented using a graph cut algorithm.

[0009] In combination with the first aspect, the two-dimensional laser point cloud depth image obtained by laser radar spherical projection includes: The Cartesian three-dimensional coordinates of the point cloud of the target area are converted into spherical three-dimensional coordinates using the following formula: , , in, is the Cartesian three-dimensional coordinate of the point cloud of the target area, , and is the pitch angle and yaw angle of the point cloud of the target area in spherical three-dimensional coordinates; The spherical three-dimensional coordinates of the point cloud of the target area are converted into two-dimensional image coordinates using the following formula: , , Among them, among them, is the maximum elevation angle of the lidar, is the row coordinate of the point cloud in the two-dimensional laser point cloud depth image, is the column coordinate of the point cloud in the two-dimensional laser point cloud depth image.

[0010] In combination with the first aspect, segmenting the point cloud information into non-ground points, unknown classification points, and ground points includes: Divide the two-dimensional laser point cloud depth image into grids of equal size, and take each grid and its 3×3 adjacent grids as a unit; Record the minimum point cloud height in each grid and the minimum point cloud height in its corresponding unit ; Set the height threshold interval , and segment the point cloud information according to the following formula: , in, is the ground point, is a non-ground point, is an unknown classification point, is the actual measured point cloud height.

[0011] In combination with the first aspect, constructing a graph model based on the marked two-dimensional laser point cloud depth image based on the conditional random field includes: Extracting segmentation features from the two-dimensional laser point cloud depth image, wherein the segmentation features include laser point cloud distance, edge information, and laser intensity; Calculate the node energy and edge energy of each pixel in the two-dimensional laser point cloud depth map according to the segmentation feature; A graph model is constructed according to the node energy and edge energy.

[0012] In combination with the first aspect, the node energy of each pixel is calculated by the following formula: , Among them, It's pixels Tags, It's pixels characteristics.

[0013] In combination with the first aspect, the edge energy of each pixel is calculated by the following formula: , in and It's pixels and Features, is a parameter that controls the similarity.

[0014] In combination with the first aspect, the performing ground segmentation on the graph model by using a graph cut algorithm includes: According to the graph model, a function to minimize the total energy is established: , in, is a tag collection, is the weight parameter of edge energy; The minimum total energy function is solved according to the minimum cut algorithm to obtain the point cloud of the ground points and non-ground points.

[0015] In combination with the first aspect, solving the minimized total energy function according to the graph cut algorithm to obtain the ground point and non-ground point point cloud includes: Initialize the information amount of each pixel in the two-dimensional laser point cloud depth map; Get the path from the source to the sink; The two-dimensional laser point cloud depth map is divided into a portion reachable from the source point and an inaccessible portion according to the path, so as to obtain the ground point and non-ground point point clouds.

[0016] According to a second aspect of the present disclosure, there is provided an electronic device, comprising: one or more processors; A storage unit is used to store one or more programs, which, when executed by the one or more processors, enable the one or more processors to implement any of the above-mentioned laser radar ground segmentation algorithms.

[0017] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that when the computer program is executed by a processor, it can implement any of the above-mentioned laser radar ground segmentation algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of the laser radar ground segmentation algorithm of the embodiment of the present disclosure; Figure 2 It is a schematic diagram of a spherical projection model of a laser radar according to an embodiment of the present disclosure; Figure 3 is a schematic diagram of a pre-classification model according to an embodiment of the present disclosure; Figure 4 It is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0019] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the embodiments of the present disclosure.

[0020] The terms used in the disclosed embodiments are only for the purpose of describing specific embodiments and are not intended to limit the disclosed embodiments. The singular forms of "a", "said" and "the" used in the disclosed embodiments and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0021] It should be understood that although the terms first, second, third, etc. may be used to describe various information in the disclosed embodiments, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the disclosed embodiments, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0022] like Figure 1 FIG. 1 is a flow chart of the laser radar ground segmentation algorithm of the embodiment of the present disclosure. The segmentation algorithm includes the following steps: S101: acquiring point cloud information of a target area through a laser radar, and obtaining a two-dimensional laser point cloud depth image through spherical projection of the laser radar, wherein each pixel of the two-dimensional laser point cloud depth image corresponds to a point cloud; S102: Segmenting the point cloud information into non-ground points, unknown classification points, and ground points according to the two-dimensional laser point cloud depth image; S103: marking the non-ground points, unknown classification points, and ground points in the two-dimensional laser point cloud depth image; S104: constructing a graph model for the marked two-dimensional laser point cloud depth image based on a conditional random field; S105: Perform ground segmentation on the graph model using a graph cut algorithm.

[0023] Refer to the following Figure 2-Figure 4 The segmentation algorithm is described in detail.

[0024] The point cloud information of the target area is acquired by the laser radar, and a two-dimensional laser point cloud depth image is obtained by spherical projection of the laser radar, wherein each pixel of the two-dimensional laser point cloud depth image corresponds to a point cloud.

[0025] Specifically, the LiDAR obtains information about the surrounding environment by emitting lasers and measuring the reflection time. By continuously rotating and scanning, the LiDAR generates a 3D point cloud data set, each point has three coordinates. , indicating the position of the point in three-dimensional space.

[0026] Furthermore, the above point cloud is projected onto a two-dimensional plane. The spherical projection process of the laser radar is as follows: Spherical coordinate system: Each three-dimensional point Convert to spherical coordinates ,in is the distance (depth), and are the horizontal and vertical angles.

[0027] By using the horizontal angle and vertical angle , convert spherical coordinates into coordinates on a two-dimensional image .here, and Corresponding to the horizontal and vertical coordinates of the image, respectively, and the depth information As the pixel values ​​of the image.

[0028] The 2D laser point cloud depth image obtained in this way can be used for various image processing tasks, such as image segmentation, target detection and recognition, etc. The depth relationship and features between pixels can be used for more accurate image processing and analysis.

[0029] Specific (such as Figure 2 ), convert the Cartesian three-dimensional coordinates of the point cloud of the target area into spherical three-dimensional coordinates using the following formula: , , in, is the Cartesian three-dimensional coordinate of the point cloud of the target area, , and is the pitch angle and yaw angle of the point cloud of the target area in spherical three-dimensional coordinates; The spherical three-dimensional coordinates of the point cloud of the target area are converted into two-dimensional image coordinates using the following formula: , , Among them, among them, is the maximum elevation angle of the lidar, is the row coordinate of the point cloud in the two-dimensional laser point cloud depth image, is the column coordinate of the point cloud in the two-dimensional laser point cloud depth image.

[0030] The point cloud information is segmented into non-ground points, unknown classification points, and ground points according to the two-dimensional laser point cloud depth image.

[0031] In the image segmentation task, each pixel is classified into the following categories: Ground point: represents a point on the ground.

[0032] Non-ground points: represent objects above the ground, such as buildings, trees, vehicles, etc.

[0033] Unknown classification point: indicates a point whose classification cannot be determined.

[0034] Specifically, the entire 2D depth image is divided into multiple small grids of equal size. These grids are helpful for local analysis of point cloud data. During processing, each grid and its adjacent 3×3 grids are regarded as a unit.

[0035] For each grid, record the following two values: : The minimum point cloud height in the grid.

[0036] : The minimum point cloud height of all points in the grid and its 3×3 adjacent grids.

[0037] Set a height threshold interval It should be noted that the threshold interval can be set by those skilled in the art according to actual needs, and this embodiment does not make any specific limitation here.

[0038] According to the following rules, the point cloud information is divided into three categories: Ground point ( ): If the difference between the current point cloud height and the minimum height of the point cloud contained in the grid is less than the threshold, the point is classified as a ground point.

[0039] Non-ground points ( ): If the difference between the current point cloud height and the minimum height of the point cloud contained in the grid is greater than the threshold, or the difference between the current point cloud height and the minimum height of the point cloud contained in the corresponding cell is greater than the threshold, the point is classified as a non-ground point.

[0040] Unknown classification point ( ): In other cases, the point is classified as an unknown classification point.

[0041] like Figure 3 , which is a schematic diagram of the pre-classification model of an embodiment of the present disclosure, wherein the top layer is non-ground points, the middle layer is unknown classification points, and the bottom layer is ground points.

[0042] Furthermore, the non-ground points, unknown classification points, and ground points are marked in the two-dimensional laser point cloud depth image. Specifically, it includes: Prepare a labeled image: Create a labeled image of the same size as the 2D laser point cloud depth image. This image is used to display the classification results of each pixel.

[0043] Assign colors: assign different colors or marks to each category: Ground points: can be marked with green, blue or other easily identifiable colors. Non-ground points: can be marked with red or other contrasting colors. Unknown classification points: can be marked with gray, black or other neutral colors. This embodiment does not make specific restrictions.

[0044] Update the marker image: For each pixel, apply the corresponding color or label to the corresponding position in the marker image based on the previous classification result.

[0045] Exemplarily, if a certain pixel is marked as a ground point in the segmentation result, the position of the pixel in the marked image is filled with the color corresponding to the ground point.

[0046] Overlaying the labeled image with the original 2D laser point cloud depth image or displaying it separately can help understand and verify the segmentation results. The labeled image can clearly show the classification of each point.

[0047] A graph model is constructed based on the labeled two-dimensional laser point cloud depth image based on the conditional random field.

[0048] Conditional random fields are a probabilistic graph model for processing labeled sequence data. It can effectively take into account the dependencies between labels and the mutual influence between pixels, thereby improving the accuracy of image segmentation. Conditional random fields mainly include node energy and edge energy.

[0049] Specifically, first of all, In the 2D laser point cloud depth image, the following segmentation features are extracted: Laser point cloud distance: The depth value of each pixel represents the distance from the laser radar to that point.

[0050] Edge information: Image edges extracted by edge detection algorithms help identify the boundaries of objects.

[0051] Laser intensity: The intensity of the laser radar return signal, which reflects the reflective characteristics of the object surface.

[0052] Then, the node energy and edge energy of each pixel in the two-dimensional laser point cloud depth map are calculated according to the segmentation features:

[0053] Among them, It's pixels Tags, It's pixels characteristics.

[0054]

[0055] in and It's pixels and Features, is a parameter that controls the similarity.

[0056] Nodes represent pixels, edges represent the neighboring relationship between pixels, and node energy is estimated by a probability model. Node energy measures the possibility of each pixel belonging to a ground point or a non-ground point. Edge energy measures the similarity between adjacent pixels.

[0057] Finally, the graph model is constructed based on node energy and edge energy: Define the structure of the graph: Node: Each pixel is a node in the graph. Edge: Define the neighborhood of each node (such as 4-neighborhood or 8-neighborhood) and establish edges between nodes.

[0058] Allocate node energy to each node: Allocate the calculated node energy value to the corresponding node.

[0059] Assign edge energy to each pair of adjacent nodes: Assign the calculated edge energy values ​​to the edges connecting adjacent nodes.

[0060] The graph model is ground segmented using a graph cut algorithm.

[0061] Specifically, first of all, Constructing the total energy function: The total energy function combines all node energies and edge energies to describe the segmentation cost of the entire image.

[0062] The formula is as follows: , in, is a tag collection, is the weight parameter of edge energy.

[0063] Then, use a graph cut algorithm, etc. to minimize the total energy function: Initialize the information amount of each pixel in the two-dimensional laser point cloud depth map; Get the path from the source to the sink; The two-dimensional laser point cloud depth map is divided into a portion reachable from the source point and an inaccessible portion according to the path, so as to obtain the ground point and non-ground point point clouds.

[0064] Specifically, source point: represents the ground point area in the image.

[0065] Sink points: represent non-ground point areas in the image.

[0066] Edge weight: The edge weight in the graph is calculated from the node energy and edge energy. The edge weight from the source to each node is related to the node energy, and the edge weight from the sink to each node is also related to the node energy.

[0067] Initialize flow: Set the initial flow of all edges in the network to 0. The flow represents the label assignment probability of each pixel.

[0068] Search path: Use the maximum flow algorithm to find an augmenting path from the source to the sink. An augmenting path is a path where the flow can be increased.

[0069] Update flow: Increase flow along the found paths until the flow of all paths reaches the maximum value.

[0070] Segment the image: The image segmentation result is determined by finding the minimum cut from the source to the sink. The minimum cut is an operation that divides the image into two parts, one of which contains the source points and the other contains the sink points.

[0071] Part reachable from source point: The part connected to the source point is marked as a ground point.

[0072] Part reachable from the sink: The part connected to the sink is marked as a non-ground point.

[0073] The embodiments of the present disclosure provide a laser radar ground segmentation algorithm, electronic device and storage medium, which effectively improves the accuracy and real-time performance of ground segmentation through two-dimensional projection and elevation pre-classification of laser point clouds, combined with conditional random field models and graph cut algorithms, reduces the amount of calculation and parameter adjustment requirements, and thus provides more reliable perception, positioning and navigation support in different environments. This innovative method overcomes the limitations of traditional methods under the framework of combining preliminary rough segmentation with fine segmentation, and improves the adaptability and performance of autonomous driving systems in complex environments.

[0074] The electronic device 400 may be a desktop computer, a notebook, a PDA, a cloud server, or other electronic device. The electronic device 400 may include, but is not limited to, a processor 401 and a memory 402. Those skilled in the art will appreciate that Figure 4 It is only an example of the electronic device 400 and does not constitute a limitation of the electronic device 400. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0075] The processor 401 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0076] The memory 402 may be an internal storage unit of the electronic device 400, for example, a hard disk or memory of the electronic device 400. The memory 402 may also be an external storage device of the electronic device 400, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 400. Further, the memory 402 may also include both an internal storage unit of the electronic device 400 and an external storage device. The memory 402 is used to store the computer program 403 and other programs and data required by the electronic device. The memory 402 may also be used to temporarily store data that has been output or is to be output.

[0077] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present disclosure. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0078] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0079] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example 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 performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.

[0080] In the embodiments provided in the present disclosure, it should be understood that the disclosed / electronic devices and methods can be implemented in other ways. For example, the / electronic device embodiments described above are only schematic, for example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation, multiple units or components may be combined or may be integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection of units, which may be electrical, mechanical or other forms.

[0081] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0082] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0083] If the integrated module / unit is implemented in the form of 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, the present disclosure implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. The computer program may include computer program code, and the computer program code may be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or recording medium that can carry computer program code, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electric carrier signals and telecommunication signals.

[0084] The above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should all be included in the protection scope of the present disclosure.

Claims

1. A laser radar ground segmentation algorithm, characterized in that: The segmentation algorithm includes: The point cloud information of the target area is obtained by laser radar, and a two-dimensional laser point cloud depth image is obtained by laser radar spherical projection, wherein each pixel of the two-dimensional laser point cloud depth image corresponds to a point cloud; Segmenting the point cloud information into non-ground points, unknown classification points, and ground points according to the two-dimensional laser point cloud depth image; Marking the non-ground points, unknown classification points, and ground points in the two-dimensional laser point cloud depth image; Constructing a graph model based on the labeled two-dimensional laser point cloud depth image based on the conditional random field; The graph model is ground segmented using a graph cut algorithm.

2. The segmentation algorithm according to claim 1, characterized in that The two-dimensional laser point cloud depth image obtained by laser radar spherical projection includes: The Cartesian three-dimensional coordinates of the point cloud of the target area are converted into spherical three-dimensional coordinates using the following formula: , , in, is the Cartesian three-dimensional coordinate of the point cloud of the target area, , and is the pitch angle and yaw angle of the point cloud of the target area in spherical three-dimensional coordinates; The spherical three-dimensional coordinates of the point cloud of the target area are converted into two-dimensional image coordinates using the following formula: , , Among them, among them, is the maximum elevation angle of the lidar, is the row coordinate of the point cloud in the two-dimensional laser point cloud depth image, is the column coordinate of the point cloud in the two-dimensional laser point cloud depth image.

3. The segmentation algorithm according to claim 1, characterized in that: The segmenting of the point cloud information into non-ground points, unknown classification points, and ground points comprises: Divide the two-dimensional laser point cloud depth image into grids of equal size, and take each grid and its 3×3 adjacent grids as a unit; Record the minimum point cloud height in each grid and the minimum point cloud height in its corresponding unit ; Set the height threshold interval , and segment the point cloud information according to the following formula: , in, is the ground point, is a non-ground point, is an unknown classification point, is the actual measured point cloud height.

4. The segmentation algorithm according to claim 1, characterized in that: The method of constructing a graph model based on the marked two-dimensional laser point cloud depth image based on the conditional random field includes: Extracting segmentation features from the two-dimensional laser point cloud depth image, wherein the segmentation features include laser point cloud distance, edge information, and laser intensity; Calculate the node energy and edge energy of each pixel in the two-dimensional laser point cloud depth map according to the segmentation feature; A graph model is constructed according to the node energy and edge energy.

5. The segmentation algorithm according to claim 4, characterized in that: The node energy of each pixel is calculated by the following formula: , Among them, It's pixels Tags, It's pixels characteristics.

6. The segmentation algorithm according to claim 4, characterized in that: The edge energy of each pixel is calculated by the following formula: , in and It's pixels and Features, is a parameter that controls the similarity.

7. The segmentation algorithm according to claim 1, characterized in that: The ground segmentation of the graph model by using a graph cut algorithm comprises: According to the graph model, a function to minimize the total energy is established: , in, is a tag collection, is the weight parameter of edge energy; The minimum total energy function is solved according to the graph cut algorithm to obtain the point cloud of the ground points and non-ground points.

8. The segmentation algorithm according to claim 1, characterized in that: Solving the minimized total energy function according to the minimum cut algorithm to obtain the ground point and non-ground point cloud includes: Initialize the information amount of each pixel in the two-dimensional laser point cloud depth map; Get the path from the source to the sink; The two-dimensional laser point cloud depth map is divided into a portion reachable from the source point and an inaccessible portion according to the path, so as to obtain the ground point and non-ground point point clouds.

9. An electronic device, characterized in that: include: one or more processors; A storage unit, used to store one or more programs, which, when executed by the one or more processors, enable the one or more processors to implement the laser radar ground segmentation algorithm according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it can implement the laser radar ground segmentation algorithm according to any one of claims 1 to 8.

Citation Information

Patent Citations

  • Three-dimensional laser radar point cloud target segmentation method based on depth map

    CN109961440A

Cited By

  • Vehicle positioning system and method based on field end non-repetitive scanning laser radar

    CN120577783A