A roadside extraction method and device

Abstract

The present invention discloses a curb extraction method, which relates to the field of unmanned driving technology. The method comprises acquiring raw point cloud data of a road environment; preprocessing a single frame of point cloud data; selecting point cloud data obtained by scanning lines that scan the curb and whose distance from the vehicle is less than a threshold; establishing a rectangular coordinate system with the vehicle's forward direction as the x-axis, selecting points with a horizontal coordinate greater than 0, and then classifying the points into right curb candidate points and left curb candidate points based on whether the vertical coordinate is greater than 0, and sorting the selected points from smallest to largest direction angle; selecting curb feature points based on vertical height difference and horizontal distance thresholds; converting the curb feature points to a map coordinate system; and generating a curb line for the corresponding side based on the left and right curb points extracted from a number of consecutive frames of point cloud data. The method can effectively avoid the influence of obstacles on curb point extraction, and the generated curb line is highly accurate.

Description

Technical Field

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

[0002] For autonomous driving systems and advanced driver assistance systems, the environmental perception system serves as the interface between the intelligent vehicle platform and the surrounding traffic scene. It also serves as the front-end input for the intelligent vehicle platform's motion decision-making and planning control systems. Road edge recognition is a crucial component of the environmental perception system.

[0003] Existing curb extraction methods struggle with obstacles blocking the curb. For example, if there are vehicles or pedestrians blocking the curb on either side of the acquisition vehicle, the extracted curb points may fall on the vehicles or pedestrians, resulting in inaccurate extraction. Furthermore, when extracting curved curbs at intersections, the extracted curb points may fall outside the curb. Furthermore, when the acquisition vehicle changes lanes, the large angle between the vehicle's forward direction and the curb can also cause the extracted curb points to fall outside the curb. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a curb extraction method and device.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0006] A curb extraction method, comprising:

[0007] Acquire raw point cloud data of a road environment obtained by a laser radar scan during vehicle travel, wherein the raw point cloud data includes a plurality of continuous frames of point cloud data;

[0008] Pre-process single-frame point cloud data to remove noise and invalid points, and use deep learning methods to detect and remove dynamic targets in the point cloud data;

[0009] Filtering point cloud data obtained by scanning lines that scan the roadside and whose distance from the vehicle is less than a threshold from the single-frame point cloud data;

[0010] A rectangular coordinate system is established with the vehicle's forward direction as the x-axis. Points with abscissas greater than 0 are selected from the filtered point cloud data. Based on whether the ordinates are greater than 0, the points are divided into right-hand curb candidate points and left-hand curb candidate points. The selected points are sorted from smallest to largest direction angle.

[0011] Based on the feature point height threshold, the right curb feature point and the left curb feature point are respectively screened from the right curb candidate point and the left curb candidate point, and it is determined whether the distance between any feature point and its adjacent feature point is greater than the horizontal distance threshold. If so, the feature point is removed, thereby extracting the left curb point and the right curb point in the frame point cloud data;

[0012] Convert the extracted left and right curb points from the lidar coordinate system to the map coordinate system;

[0013] The roadside line on the corresponding side is generated based on the left roadside points and the right roadside points extracted from several consecutive frames of point cloud data.

[0014] As a preferred solution of the curb extraction method of the present invention, the preprocessing of the single-frame point cloud data to remove noise points and invalid points includes:

[0015] Set the effective point height threshold and remove the point cloud data that is higher than the effective point height threshold;

[0016] A valid point cloud area is set with the vehicle location as the center, and point cloud data outside the valid point cloud area is removed.

[0017] As a preferred solution of the curb extraction method described in the present invention, the effective point cloud area is a rectangular area with the location of the laser radar as the coordinate origin and Xmin, Xmax, Ymin, and Ymax as endpoints.

[0018] As a preferred solution of the curb extraction method of the present invention, wherein: the sorting of the selected points according to the direction angle from small to large includes:

[0019] By formula Calculate the angle between the point and the x-axis, where a is the direction angle, x is the horizontal coordinate of the point, and |y| is the absolute value of the vertical coordinate of the point;

[0020] The angle between the point and the x-axis is taken as the direction angle of the point;

[0021] Sort the filtered points according to the direction angle from small to large.

[0022] As a preferred embodiment of the curb extraction method of the present invention, the step of generating a curb line on the corresponding side based on the left curb points and the right curb points extracted from a plurality of consecutive frames of point cloud data includes:

[0023] Cluster the left curb points and right curb points extracted from several consecutive frames of point cloud data respectively;

[0024] The clustered point sequence is downsampled using the mean sampling method;

[0025] The curb points on the corresponding side are sorted according to the direction of the curb to generate the curb line on the corresponding side.

[0026] As a preferred solution of the curb extraction method of the present invention, the height threshold of the feature point is 10 to 20 cm.

[0027] The present invention also discloses a curb extraction device, comprising:

[0028] An acquisition module is used to acquire raw point cloud data of the road environment obtained by the laser radar scanning during the vehicle's driving process, wherein the raw point cloud data includes a plurality of frames of continuous point cloud data;

[0029] The preprocessing module is used to preprocess the single-frame point cloud data and remove noise and invalid points;

[0030] The first screening module is used to screen out point cloud data obtained by scanning a scanning line that scans the roadside and has a distance from the vehicle less than a threshold from the single-frame point cloud data;

[0031] A second screening module is configured to establish a rectangular coordinate system with the vehicle's forward direction as the x-axis, further screen out points with a horizontal coordinate greater than 0 from the point cloud data screened by the first screening module, and divide the points into right curb candidate points and left curb candidate points based on whether the vertical coordinate is greater than 0;

[0032] A third screening module is used to screen out right curb feature points and left curb feature points from the right curb candidate points and the left curb candidate points respectively based on the feature point height threshold, and to determine whether there is any feature point whose distance to its adjacent feature point is greater than the horizontal distance threshold, and to remove the feature point if such a feature point exists, so as to extract the left curb point and the right curb point in the frame point cloud data;

[0033] The generation module is used to generate the roadside line on the corresponding side according to the left roadside points and the right roadside points extracted from a plurality of consecutive frames of point cloud data.

[0034] As a preferred embodiment of the curb extraction device of the present invention, the pre-processing module is used to:

[0035] Set the effective point height threshold and remove the point cloud data that is higher than the effective point height threshold;

[0036] A valid point cloud area is set with the vehicle location as the center, and point cloud data outside the valid point cloud area is eliminated;

[0037] Dynamic targets in point cloud data are detected and removed through deep learning methods.

[0038] The present invention also discloses a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the roadside extraction method as described in any of the above schemes is implemented.

[0039] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the roadside extraction method as described in any of the above schemes is implemented.

[0040] The beneficial effects of the present invention are:

[0041] The present invention detects and eliminates dynamic targets in point cloud data through a deep learning method, which can effectively avoid the influence of obstacles on the road on the extraction of curb points. At the same time, after extracting the right curb candidate points and the left curb candidate points, these points are sorted from small to large according to the direction angle of the points, which effectively avoids the problem of inaccurate sorting results of the extracted curb points when there is a curved curb, resulting in incorrect extraction of curb points, ensuring the accuracy of the generated curb line, and being able to improve the quality and mapping efficiency of high-precision maps. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0043] Figure 1 A schematic diagram of the process of the curb extraction method provided by the present invention;

[0044] Figure 2 This is a schematic diagram of the specific process of step S102 in the curb extraction method provided by the present invention;

[0045] Figure 3 A schematic diagram of a specific process of sorting points according to their direction angles from small to large in the curb extraction method provided by the present invention;

[0046] Figure 4 A schematic diagram of a specific process for generating a roadside line in the roadside extraction method provided by the present invention;

[0047] Figure 5 A schematic structural diagram of the curb extraction device provided by the present invention;

[0048] Figure 6 This is a schematic diagram of the extraction effect of curb points on a straight road section;

[0049] Figure 7 This is a schematic diagram of the extraction effect of the curb points at the intersection;

[0050] Among them: 110, acquisition module; 120, pre-processing module; 130, first screening module; 140, second screening module; 150, third screening module; 160, generation module. DETAILED DESCRIPTION

[0051] In order to make the contents of the present invention more clearly understood, the present invention is further described below in detail based on specific implementation methods in conjunction with the accompanying drawings.

[0052] See also Figure 1 , is a flow chart of the curb extraction method provided in this embodiment. The method specifically includes steps S101 to S105, and the specific steps are described as follows:

[0053] Step S101: Acquire original point cloud data of the road environment obtained by laser radar scanning during vehicle driving.

[0054] Specifically, while the vehicle is driving on the road, the laser radar continuously scans and collects the road environment around the vehicle to obtain the original point cloud data of the road section. It can be understood that the original point cloud data includes several frames of continuous point cloud data.

[0055] Step S102: Pre-process the single-frame point cloud data to remove noise and invalid points, and detect dynamic targets in the point cloud data through a deep learning method and remove them.

[0056] The raw point cloud data obtained by LiDAR scanning contains many noise points and invalid points, so the raw point cloud data needs to be preprocessed first. The preprocessing operations for single-frame point cloud data include:

[0057] Step S102a: Set a valid point height threshold and remove point cloud data that is higher than the valid point height threshold. The valid point height threshold can be used to remove noise points that are higher than the road surface.

[0058] Step S102b: Define a valid point cloud region and remove point cloud data outside the valid point cloud region. The valid point cloud region is defined as a rectangular region with the lidar location as the coordinate origin and Xmin, Xmax, Ymin, and Ymax as endpoints. Xmin and Xmax are the two endpoints of the point cloud region on the horizontal axis, and Ymin and Ymax are the two endpoints of the point cloud region on the vertical axis. In this embodiment, the values ​​of Xmin, Xmax, Ymin, and Ymax are: -10m, 30m, -20m, and 20m, respectively.

[0059] Afterwards, dynamic objects in the point cloud data are detected and removed using deep learning methods. The deep learning 3D object detection model Pointpillars detects dynamic objects such as people and vehicles in the point cloud data and filters out these invalid points based on the output 3D detection bounding boxes.

[0060] Step S103: Filtering out point cloud data obtained by scanning a scanning line that scans the roadside and has a distance from the vehicle less than a threshold from the single-frame point cloud data.

[0061] The laser radar's scanning beams are concentric circles, from which several scanning beams (i.e., scanning lines) that can scan the curb and are close to the vehicle are selected, and then the point cloud data obtained by scanning these scanning lines is selected. Referring to the figure, in this embodiment, six scanning lines are selected.

[0062] Step S104: Establish a rectangular coordinate system with the vehicle's forward direction as the x-axis, filter out points with horizontal coordinates greater than 0 from the filtered point cloud data, and then divide the points into right-side curb candidate points and left-side curb candidate points based on whether the vertical coordinate is greater than 0, and sort the filtered points from small to large according to the direction angle.

[0063] Specifically, a rectangular coordinate system is established with the vehicle's location as the origin, the vehicle's forward direction as the x-axis, and the horizontal plane perpendicular to the vehicle's forward direction as the y-axis. The point cloud data filtered in step S103 is then filtered again based on points with abscissas greater than 0, i.e., points located in front of the vehicle. Among these filtered points, those with ordinates greater than 0 are classified as right-hand curb candidate points, while those with ordinates less than 0 are classified as left-hand curb candidate points.

[0064] After extracting the right and left roadside candidate points, sort these points in ascending order of their direction angles. The specific steps include:

[0065] Step S104a: By formula Calculates the angle between a point and the x-axis, where a is the direction angle, x is the horizontal coordinate of the point, and |y| is the absolute value of the vertical coordinate of the point.

[0066] Step S104b: taking the angle between the point and the x-axis as the direction angle of the point.

[0067] Step S104c: Sort the selected points in ascending order of direction angle.

[0068] Step S105: Based on the feature point height threshold, the right side curb feature points and the left side curb feature points are respectively selected from the right side curb candidate points and the left side curb candidate points, and it is determined whether the distance between any feature point and its adjacent feature point is greater than the horizontal distance threshold. If so, the feature point is eliminated, thereby extracting the left side curb point and the right side curb point in the frame point cloud data.

[0069] Specifically, for the sorted point cloud data, the roadside feature points are screened using the two features of vertical height difference and horizontal distance threshold. The roadside is higher than the road surface, so the roadside feature points can be roughly screened out based on the feature point height threshold. In addition, the roadside points are continuous in space. If the distance between the current point and the adjacent point is too large, exceeding the horizontal distance threshold, it may be that a vehicle has passed through the point cloud data obtained by the current scan line and has not been completely filtered in the preprocessing of step S102. Therefore, this point cannot be used as a candidate roadside point and needs to be eliminated. The points extracted after screening by the two features of vertical height difference and horizontal distance threshold are the left roadside point and the right roadside point in the point cloud data of this frame.

[0070] Generally, the curb is 10 to 20 cm higher than the road surface. In this embodiment, the feature point height threshold is set to 15 cm. The horizontal distance threshold depends on the model of the laser radar.

[0071] Step S106: converting the extracted left curb point and right curb point from the lidar coordinate system to the map coordinate system.

[0072] Specifically, the approximate pose information is obtained based on the timestamp of the current lidar data, and the pose information at the current time is calculated through linear interpolation. The point cloud coordinate system is converted to the vehicle's baselink coordinate system through matrix transformation, and then the baselink coordinate system is converted to the map coordinate system.

[0073] Step S107: Generate a roadside line on the corresponding side according to the left roadside points and the right roadside points extracted from a plurality of consecutive frames of point cloud data.

[0074] Specifically, steps S102 to S106 are to extract the left curb points and the right curb points from a single frame of point cloud data. By looping steps S102 to S106, all the left curb points and the right curb points of the road section can be extracted from a number of consecutive frames of point cloud data. Then, the left curb line and the right curb line are generated based on the left curb points and the right curb points. The specific steps for generating the curb line are as follows:

[0075] Step S107a: clustering the left curb points and the right curb points extracted from a plurality of consecutive frames of point cloud data using the DBSCANE density clustering algorithm.

[0076] Step S107b: Down-sampling the clustered point sequence along the x-direction or y-direction using the mean sampling method, so that the distribution of the output roadside points after sampling will be more uniform.

[0077] Step S107c: Sort the corresponding side curb points according to the direction of the curb to generate the corresponding side curb line. The direction of the curb is determined by the slope k of the fitted line after clustering. If |k|>=1, the points are sorted from small to large along the Y axis; if |k|<1, the points are sorted along the X axis.

[0078] Therefore, the above-mentioned roadside extraction method can effectively avoid the influence of obstacles on the extraction of roadside points, and the generated roadside lines are highly accurate, which can improve the quality and mapping efficiency of high-precision maps.

[0079] See also Figure 5 , a curb extraction device provided in this embodiment includes an acquisition module 110, a preprocessing module 120, a first screening module 130, a second screening module 140, a third screening module 150 and a generation module 160.

[0080] The acquisition module 110 is used to acquire the original point cloud data of the road environment obtained by the laser radar scanning during the vehicle's driving process. The original point cloud data includes several frames of continuous point cloud data.

[0081] The pre-processing module 120 is used to pre-process the single-frame point cloud data and remove noise points and invalid points.

[0082] In another embodiment, the preprocessing module 120 is used to set a valid point height threshold and eliminate point cloud data that is higher than the valid point height threshold; set a valid point cloud area centered on the vehicle's location and eliminate point cloud data outside the valid point cloud area; detect dynamic targets in the point cloud data through a deep learning method and eliminate them.

[0083] The first screening module 130 is used to screen out point cloud data obtained by scanning a scanning line that scans the roadside and has a distance from the vehicle less than a threshold from the single-frame point cloud data.

[0084] The second screening module 140 is used to establish a rectangular coordinate system with the vehicle's forward direction as the x-axis, further screen out points with horizontal coordinates greater than 0 from the point cloud data screened out by the first screening module, and divide the points into right-side curb candidate points and left-side curb candidate points according to whether the vertical coordinate is greater than 0.

[0085] The third screening module 150 is used to screen out right-side curb feature points and left-side curb feature points from the right-side curb candidate points and the left-side curb candidate points respectively based on the feature point height threshold, and to determine whether there is any feature point whose distance to its adjacent feature point is greater than the horizontal distance threshold, and to remove the feature point if it exists, so as to extract the left-side curb point and the right-side curb point in the frame point cloud data.

[0086] The generation module 160 is used to generate a roadside line on the corresponding side according to the left roadside points and the right roadside points extracted from a plurality of consecutive frames of point cloud data.

[0087] In another embodiment, the generation module 160 is used to cluster the left curb points and the right curb points extracted from several consecutive frames of point cloud data using the DBSCANE density clustering algorithm; downsample the clustered point sequence along the x direction or the y direction using the mean sampling method; sort the corresponding side curb points according to the direction of the curb to generate the curb line on the corresponding side.

[0088] This embodiment further provides a computer device, the components of which may include but are not limited to: one or more processors or processing units, a system memory, and a bus connecting different system components (including the system memory and the processing unit).

[0089] The term "bus" refers to one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0090] The computer system / server typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer system / server, including volatile and non-volatile media, removable and non-removable media.

[0091] The system memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The computer device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media. A disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") may be provided, as well as an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM or other optical medium). In these cases, each drive may be connected to the bus via one or more data medium interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of an embodiment of the present invention.

[0092] A program / utility having a set (at least one) of program modules, which may be stored, for example, in a memory, includes, but is not limited to, an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may include an implementation of a network environment. The program modules generally implement the functions and / or methods of the embodiments described herein.

[0093] A computer device may also communicate with one or more external devices, such as a keyboard, pointing device, display, etc. Such communication may be performed via an input / output (I / O) interface. Furthermore, a computer device may also communicate with one or more networks, such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet, via a network adapter.

[0094] The processing unit executes the functions and / or methods described in the embodiments of the present invention by running the programs stored in the system memory.

[0095] The above-mentioned computer program can be set in a computer storage medium, that is, the computer storage medium is encoded with a computer program, and when the program is executed by one or more computers, it enables one or more computers to perform the method flow and / or device operation shown in the above-mentioned embodiments of the present invention.

[0096] As time goes by and technology develops, the meaning of medium becomes more and more extensive, and the propagation path of computer programs is no longer limited to tangible media, but can also be downloaded directly from the Internet. Any combination of one or more computer-readable media can be used. Computer-readable media can be computer-readable signal media or computer-readable storage media. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or components, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, device or device.

[0097] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0098] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0099] Computer program code for performing the operations of the present invention may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0100] In addition to the above embodiments, the present invention may also have other implementation methods; any technical solutions formed by equivalent replacement or equivalent transformation fall within the protection scope required by the present invention.

Claims

1. A curb extraction method, characterized by: include, Acquire raw point cloud data of a road environment obtained by a laser radar scan during vehicle travel, wherein the raw point cloud data includes a plurality of continuous frames of point cloud data; Pre-process single-frame point cloud data to remove noise and invalid points, and use deep learning methods to detect and remove dynamic targets in the point cloud data; Filtering point cloud data obtained by scanning a scanning line that scans the roadside and has a distance from the vehicle less than a threshold from the single-frame point cloud data; A rectangular coordinate system is established with the vehicle's forward direction as the x-axis and the y-axis perpendicular to the vehicle's forward direction in the horizontal plane as the y-axis. Points with abscissas greater than 0 are selected from the filtered point cloud data. The points are then divided into right-hand curb candidate points and left-hand curb candidate points based on whether their ordinates are greater than 0. The angle between the point and the x-axis is used as the point's azimuth angle, and the selected points are sorted in ascending order of azimuth angle. Based on the feature point height threshold, the right curb feature point and the left curb feature point are respectively screened from the right curb candidate point and the left curb candidate point, and it is determined whether the distance between any feature point and its adjacent feature point is greater than the horizontal distance threshold. If so, the feature point is removed, thereby extracting the left curb point and the right curb point in the frame point cloud data; Convert the extracted left and right curb points from the lidar coordinate system to the map coordinate system; The roadside line on the corresponding side is generated according to the left roadside points and the right roadside points extracted from several consecutive frames of point cloud data.

2. The curb extraction method according to claim 1, characterized in that: The pre-processing of the single-frame point cloud data to remove noise points and invalid points includes: Set the effective point height threshold and remove the point cloud data that is higher than the effective point height threshold; A valid point cloud area is set with the vehicle location as the center, and point cloud data outside the valid point cloud area is removed.

3. The curb extraction method according to claim 2, characterized in that: The effective point cloud area is a rectangular area with the laser radar location as the coordinate origin and Xmin, Xmax, Ymin, and Ymax as endpoints.

4. The curb extraction method according to claim 1, characterized in that: The sorting of the selected points according to the direction angle from small to large includes: By formula Calculate the angle between the point and the x-axis, where a is the direction angle, x is the horizontal coordinate of the point, and |y| is the absolute value of the vertical coordinate of the point; Sort the filtered points according to the direction angle from small to large.

5. The curb extraction method according to claim 1, characterized in that: The step of generating a roadside line on the corresponding side based on the left roadside points and the right roadside points extracted from a plurality of consecutive frames of point cloud data includes: Cluster the left curb points and right curb points extracted from several consecutive frames of point cloud data respectively; The clustered point sequence is downsampled using the mean sampling method; The curb points on the corresponding side are sorted according to the direction of the curb to generate the curb line on the corresponding side.

6. The curb extraction method according to claim 1, characterized in that: The feature point height threshold is 10 to 20 cm.

7. A curb extraction device, characterized in that: include, An acquisition module is used to acquire raw point cloud data of the road environment obtained by the laser radar scanning during the vehicle's driving process, wherein the raw point cloud data includes a plurality of frames of continuous point cloud data; The preprocessing module is used to preprocess single-frame point cloud data, remove noise and invalid points, and detect dynamic targets in point cloud data through deep learning methods and remove them; The first screening module is used to screen out point cloud data obtained by scanning a scanning line that scans the roadside and has a distance from the vehicle less than a threshold from the single-frame point cloud data; A second screening module is configured to establish a rectangular coordinate system with the vehicle's forward direction as the x-axis and the y-axis perpendicular to the vehicle's forward direction in a horizontal plane as the y-axis, further screening out points with a horizontal coordinate greater than 0 from the point cloud data screened out by the first screening module, and classifying the points into right curb candidate points and left curb candidate points based on whether the vertical coordinate is greater than 0. The angle between the point and the x-axis is used as the direction angle of the point, and the screened points are sorted in ascending order of the direction angle; A third screening module is used to screen out right curb feature points and left curb feature points from the right curb candidate points and the left curb candidate points respectively based on the feature point height threshold, and to determine whether there is any feature point whose distance to its adjacent feature point is greater than the horizontal distance threshold, and to remove the feature point if such a feature point exists, so as to extract the left curb point and the right curb point in the frame point cloud data; The generation module is used to convert the extracted left curb points and right curb points from the lidar coordinate system to the map coordinate system, and generate the curb lines on the corresponding sides based on the left curb points and right curb points extracted from several consecutive frames of point cloud data.

8. The curb extraction device according to claim 7, characterized in that: The pre-processing module is used to: Set the effective point height threshold and remove the point cloud data that is higher than the effective point height threshold; A valid point cloud area is set with the vehicle location as the center, and point cloud data outside the valid point cloud area is removed.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images