Nothing Special   »   [go: up one dir, main page]

WO2023040437A1 - 路牙确定方法、装置、设备以及存储介质 - Google Patents

路牙确定方法、装置、设备以及存储介质 Download PDF

Info

Publication number
WO2023040437A1
WO2023040437A1 PCT/CN2022/104941 CN2022104941W WO2023040437A1 WO 2023040437 A1 WO2023040437 A1 WO 2023040437A1 CN 2022104941 W CN2022104941 W CN 2022104941W WO 2023040437 A1 WO2023040437 A1 WO 2023040437A1
Authority
WO
WIPO (PCT)
Prior art keywords
point cloud
curb
grid
raster
density
Prior art date
Application number
PCT/CN2022/104941
Other languages
English (en)
French (fr)
Inventor
唐凯涛
孔旗
Original Assignee
北京京东乾石科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京京东乾石科技有限公司 filed Critical 北京京东乾石科技有限公司
Publication of WO2023040437A1 publication Critical patent/WO2023040437A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/29—Geographical information databases
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/88—Lidar systems specially adapted for specific applications
    • G01S17/93—Lidar systems specially adapted for specific applications for anti-collision purposes
    • G01S17/931—Lidar systems specially adapted for specific applications for anti-collision purposes of land vehicles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds

Definitions

  • the present disclosure relates to the field of computer technology, in particular to the field of data processing, and in particular to a method, device, equipment and storage medium for determining curbs.
  • a high-precision map is a map form different from ordinary electronic navigation maps.
  • a high-precision map is a map that serves the automatic driving system.
  • High-precision maps consist of point cloud maps and vector maps.
  • the point cloud map refers to the environmental map collected by the laser sensor. The laser scans the surrounding environment, uses the point cloud in the space to represent the environmental information, and finally expresses the environment in the form of a three-dimensional point cloud. Because the sensing range of the laser is 360 degrees, and the angle range of the pitch angle is relatively large, most of the information in the environment will be collected, such as roads, vehicles, pedestrians, trees and buildings.
  • a vector map refers to the information contained in the road surface, such as lane lines or curbs drawn on the road surface related to traffic regulations.
  • the present disclosure provides a road curb determination method, device, equipment, storage medium and computer program product.
  • a method for determining a curb including: obtaining point cloud frames collected at multiple collection points to obtain a point cloud frame sequence; determining the ground point cloud in each point cloud frame in the point cloud frame sequence; After the ground point cloud is projected to the ground, it is divided into grids to determine multiple grid images; according to the multiple grid images, the road tooth information is determined; and the road tooth information is output.
  • a road curb determination device including: a point cloud frame acquisition unit configured to acquire point cloud frames collected at multiple collection points to obtain a sequence of point cloud frames; a ground separation unit configured to Determine the ground point cloud in each point cloud frame in the point cloud frame sequence; the grid division unit is configured to perform grid division after projecting the ground point cloud to the ground, and determine multiple grid images; the information determination unit is configured The method is to determine curb information according to multiple grid images; the information output unit is configured to output curb information.
  • an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by at least one processor. Executed by a processor, so that at least one processor can execute the method described in the first aspect.
  • a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the method described in the first aspect.
  • a computer program product includes a computer program, and when executed by a processor, the computer program implements the method as described in the first aspect.
  • the curb information can be determined directly through the three-dimensional point cloud, which improves the accuracy of the curb information.
  • FIG. 1 is an exemplary system architecture diagram to which an embodiment of the present disclosure can be applied;
  • FIG. 2 is a flowchart of an embodiment of a method for determining curbs according to the present disclosure
  • FIG. 3 is a schematic diagram of an application scenario of a method for determining curbs according to the present disclosure
  • FIG. 4 is a flowchart of another embodiment of a method for determining a curb according to the present disclosure
  • Fig. 5A is a schematic diagram of the "stop circle" of the embodiment shown in Fig. 4;
  • Fig. 5B is the original density raster map of the embodiment shown in Fig. 4;
  • Fig. 5C is a schematic diagram of an improved edge operator of the embodiment shown in Fig. 4;
  • Figure 5D is an intensity grid map of the embodiment shown in Figure 4.
  • Figure 5E is a canny edge map of the embodiment shown in Figure 4.
  • Figure 5F is a height grid map of the embodiment shown in Figure 4.
  • Fig. 6 is a schematic structural diagram of an embodiment of a curb determining device according to the present disclosure.
  • FIG. 7 is a block diagram of an electronic device used to implement the method for determining a curb in an embodiment of the present disclosure.
  • FIG. 1 shows an exemplary system architecture 100 to which embodiments of the method for determining a curb or the apparatus for determining a curb of the present disclosure can be applied.
  • a system architecture 100 may include a vehicle 101 , a network 102 , a server 103 and a terminal device 104 .
  • the network 102 is used as a medium for providing a communication link among the vehicle 101 , the server 103 and the terminal device 104 .
  • Network 102 may include various connection types, such as wires, wireless communication links, or fiber optic cables, among others.
  • a laser radar may be installed on the vehicle 101, and the laser radar is used to collect point cloud data of the surrounding environment of the vehicle 101.
  • the above point cloud data may include multiple point cloud frames.
  • the vehicle 101 can send the collected point cloud data to the server 103 or the terminal device 104 through the network 102 .
  • the server 103 can process the received point cloud data to determine the curb information. And the curb information can be fed back to the vehicle 101 or the terminal device 104 .
  • the vehicle 101 can be an automatic driving vehicle, which can better automatically drive according to the high-precision map and road curb information.
  • the server 103 may be hardware or software. When the server 103 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When the server 103 is software, it may be implemented as multiple software or software modules (for example, for providing distributed services), or as a single software or software module. No specific limitation is made here.
  • the terminal device 104 can also receive point cloud data, process the point cloud data, and display the determined curb information.
  • a point cloud data processing application can be installed on the terminal device 104 to process the point cloud data.
  • the user can view the determined road curb information through the terminal device 104 .
  • the method for determining the curb may be executed by the server 103 or the terminal device 104 .
  • the device for determining the curb can be set in the server 103 or the terminal device 104 .
  • FIG. 2 shows a flow 200 of an embodiment of a method for determining a curb according to the present disclosure.
  • the curb determining method of the present embodiment comprises the following steps:
  • Step 201 acquire point cloud frames collected at multiple collection points, and obtain a sequence of point cloud frames.
  • the execution subject can obtain point cloud frames collected at multiple collection points in various ways.
  • the execution subject can obtain the point cloud frames collected by the collection vehicle at multiple collection points on the driving route.
  • Each collection point can be separated by a preset distance.
  • Each point cloud frame may include multiple point cloud points, and each point cloud point may include height information and intensity information.
  • Point cloud frames corresponding to different collection points can form a sequence of point cloud frames.
  • Step 202 determine the ground point cloud in each point cloud frame in the point cloud frame sequence.
  • the execution subject may perform ground separation on each point cloud frame to determine the ground point cloud in each point cloud frame. Specifically, the execution subject may use the point cloud points whose height value is less than the preset height threshold in each point cloud frame as the ground point cloud. Alternatively, the execution subject may use the remaining point cloud points after removing the point cloud points of the identified obstacles in each point cloud frame as the ground point cloud.
  • Step 203 after projecting the ground point cloud to the ground, perform grid division, and determine multiple grid maps.
  • each projection point can be divided into a grid with a fixed size, so that each grid can include projections of multiple point cloud points.
  • the execution subject can determine the information of each grid according to the information of each point cloud point. For example, the intensity and height information of the grid can be determined according to the intensity and height information of the point cloud points.
  • a raster map may include multiple rasters, and the information of each raster forms a raster map. Raster maps can include height rasters, density rasters, and intensity rasters.
  • the height grid map uses the median or average of the height values of all point cloud points in the grid as the height value of the grid.
  • the density raster map uses the number of all point cloud points in the grid as the density value of the grid.
  • the intensity raster map uses the median or average of the intensity values of all point cloud points in the grid as the intensity value of the grid.
  • Step 204 determine road curb information according to multiple grid maps.
  • the execution subject can determine curb information in various ways. For example, the execution subject may input each raster image into a pre-trained curb information determination model, and use the output of the model as curb information. Alternatively, the execution subject may perform image processing on each raster image to obtain road curb information. Existing image processing algorithms, such as edge detection algorithms, erosion algorithms, expansion algorithms, etc., can be used in image processing.
  • Step 205 output curb information.
  • the execution subject After the execution subject determines the curb information, it can output the above curb information. Specifically, the execution subject can output curb information to the self-driving vehicle, so that the self-driving vehicle can perform operations such as parking according to the curb information. Alternatively, the execution subject may output the curb information to the terminal screen for viewing by the user.
  • FIG. 3 shows a schematic diagram of an application scenario of the method for determining a curb according to the present disclosure.
  • the collection vehicle 301 collects point cloud data of the surrounding environment at each collection point of the driving route. Then the collected point cloud data is sent to the server 302 .
  • the server 302 can process the point cloud data to determine the curb information of the surrounding environment. Then update the curb information into the high-precision map, and send the updated high-precision map to the self-driving vehicle 303, and the self-driving vehicle 303 can park more accurately according to the updated high-precision map.
  • the method for determining the curb provided by the above-mentioned embodiments of the present disclosure can divide the point cloud frame into a grid and determine the curb information by using the grid image, thereby improving the accuracy of determining the curb.
  • FIG. 4 shows a flow 400 of another embodiment of the method for determining a curb according to the present disclosure.
  • the method of this embodiment may include the following steps:
  • Step 401 acquire point cloud frames collected at multiple collection points, and obtain a sequence of point cloud frames.
  • Step 402 for each point cloud frame in the point cloud frame sequence, perform morphological filtering on the point cloud frame to determine the candidate ground point cloud in the point cloud frame; The point cloud points whose distance between the collection points is greater than the preset distance threshold are deleted to obtain the ground point cloud.
  • the execution subject may process each point cloud frame in the point cloud frame sequence, that is, perform morphological filtering on each point cloud frame to realize ground separation.
  • the filtered point cloud is used as the candidate ground point cloud.
  • morphological filtering can include various processing, such as dilation, erosion and other processing.
  • the execution subject may delete the point cloud points whose distance between the collection points corresponding to the point cloud frame in the candidate ground point cloud is greater than a preset distance threshold, to obtain the ground point cloud. Since the curb is generally located in the ground point cloud, only the ground point cloud is processed here, which can speed up the processing.
  • the point cloud points farther away from the collection point represent the point cloud points farther away from the car body.
  • the point cloud here is sparse and less reliable, so the untrustworthy point cloud points are deleted here.
  • Step 403 according to the acquisition time corresponding to each point cloud frame in the point cloud frame sequence, divide the point cloud frame sequence to obtain a plurality of point cloud frame subsequences; The ground point cloud is spliced; the spliced ground point cloud is projected to the ground and then divided into grids to obtain multiple grids; multiple grid images are determined according to the intensity, density and height of each point cloud point.
  • the execution subject may firstly divide each point cloud frame in the point cloud frame sequence according to time. That is, the acquisition time corresponding to each point cloud frame is divided into the point cloud frame sequence every 30S to obtain multiple point cloud frame subsequences.
  • the point cloud frame sequence is divided and then processed, which can ensure the smooth progress of point cloud processing.
  • the execution subject can splice the ground point cloud in each point cloud frame in each point cloud frame subsequence respectively. During splicing, it can be spliced according to the pose estimated by GPS and inertial navigation documents.
  • the execution subject can project the spliced point cloud to the ground and perform grid division to obtain multiple grids. Then, a plurality of raster maps are determined according to the intensity, density and height of the point cloud points included in each raster.
  • each grid can be set as a square, and the side length of the square can be set according to the actual situation.
  • the executive body can determine multiple grid maps through the following steps: according to the intensity, density and height of point cloud points included in each grid, determine the Intensity, Density, and Height; Multiple raster maps are determined based on the intensity, density, and height of each raster.
  • the execution subject may determine the strength, density and height of each grid according to the strength, density and height of point cloud points included in each grid.
  • Multiple raster maps are then determined based on the intensity, density, and height of each raster.
  • the intensity grid map uses the median or average of the intensity values of all point cloud points in the grid as the intensity value of the grid.
  • the height grid map uses the median or average of the height values of all point cloud points in the grid as the height value of the grid.
  • the density grid map uses the number of all point clouds in the grid as the density value of the grid.
  • Step 404 slice each density grid image, intensity grid image and height grid image to obtain multiple density grid image slices, intensity grid image slices and height grid image slices; according to the density grid image slices, Intensity raster slices, height raster slices and pre-trained segmentation models determine the curb masks corresponding to each raster slice; according to the density raster, height raster and each curb mask, determine the road dental information.
  • the above three grid maps can basically reflect the ground information of the environment.
  • the curbs in the grid are segmented through the deep learning model.
  • the size of the grid map divided by space is very large. If it is directly read, the memory will be full, and it cannot be directly put into the model.
  • the execution subject may slice multiple raster images according to space to obtain multiple raster image slices. The size of the slice is just right to be fed into the deep learning model for detection. Then, the execution subject can input multiple raster image slices into the pre-trained segmentation model to determine the corresponding curb mask. Specifically, the execution subject can slice each density raster image, intensity raster image, and intensity raster image respectively, and then combine the obtained density raster image slices, intensity raster image slices, and intensity raster image slices into three The channel image is input to the pre-trained segmentation model to determine the corresponding curb mask.
  • each curb mask obtained corresponds to a single point cloud frame subsequence.
  • the executive body can splice the curb masks to obtain the curb mask of the entire road.
  • the execution subject can directly use the curb mask as curb information.
  • the executive body may further process the curb mask to obtain clearer and more accurate curb information.
  • the execution subject can determine the grid image slice through the following steps: in the point cloud frame subsequence corresponding to each grid image, according to the position of the collection point and the preset slice size , slice the plurality of density raster images to obtain a plurality of density raster image slices.
  • the execution subject may slice the plurality of density raster images according to the location of the collection point and a preset slice size to obtain a plurality of density raster image slices. Specifically, the execution subject can select collection points at equal intervals, frame a spatial range around the collection points, and cut them into 960*640 pixels. The size of the slices can be sent to the model for detection. This disclosure selects a model developed based on the Deeplabv3 framework.
  • the execution subject can determine the road curb information through the following steps: according to the relative position between the point cloud frame subsequences corresponding to each road curb mask, perform Splicing to obtain a spliced curb mask; according to the density raster map and the height raster map, post-process the spliced curb mask to determine the curb information.
  • the execution subject may splice each curb mask according to the relative position between the point cloud frame subsequences corresponding to each curb mask to obtain a spliced curb mask.
  • the road curb masks corresponding to adjacent point cloud frame subsequences are spliced.
  • the spliced curb mask can be obtained after splicing each curb mask.
  • the executive body can further combine the density raster map and the height raster map to perform post-processing such as fitting on the spliced curb mask to determine accurate curb information.
  • the executive body may further post-process the stitched curb mask through the following steps: refine the stitched curb mask to determine candidate curb pixels; map, determine the edge pixels; determine the edge pixels of the curb according to the candidate curb pixels and edge pixels; determine the edge information of the curb according to the height raster map and the edge pixels of the curb; fit the edge information of the curb to determine Curb information.
  • the execution subject may refine the spliced curb mask first, in order to corrode each curb into a line without width.
  • the outline of the curb can be basically seen in the thinned mask image, but the shape after the thinning algorithm is irregular, not a smooth straight line, and cannot be directly fitted, so post-processing is required.
  • the thinned pixels are referred to as curb candidate pixels.
  • the goal of post-processing is to fit the candidate curb pixels in the mosaic curb mask to geometric information, but the candidate curb pixels have width, because the actual curb also has width. And the curb vector that the target needs to output has no width.
  • the curb mask output by the above deep learning model is the edge of the curb on the side close to the road surface. This embodiment is to confirm the effective contour of the curb and ensure the safe passage of the self-driving vehicle. Before fitting, it is necessary to screen the candidate curb pixels in the spliced curb mask image, and filter out the curb on the side far away from the road surface. pixel. Only in this way can the subsequent fitting accuracy of the curb be improved.
  • the Canny edge algorithm Based on the Sobel edge algorithm, the Canny edge algorithm performs more detailed post-processing on the edge pixels, and filters the edge points around the edge point, so that the edge part is more detailed and accurate.
  • Canny edge detection can be subdivided into three steps: 1. Use Sobel convolution kernel for convolution operation. The convolution kernel of Sobel in the x direction is shown in Formula 1, and the convolution kernel of Sobel in the y direction is shown in Formula 2.
  • the overall gradient value is then calculated based on the G x and G y gradient values that yield the gradient values in the x and y directions direction angle
  • non-maximum suppression is used to select the pixel with the largest gradient value in a range, so as to make the extracted edge more stable and avoid extracting noise in the density raster image.
  • T 1 , T 2 Two thresholds T 1 , T 2 are set. Pixels with a gradient value greater than T 2 are classified as "confirmed edge pixels" and are retained. Pixels with a gradient value smaller than T 1 are considered not to belong to the edge and are discarded. Pixels whose gradient values are between T1 and T2 are considered as part of the edge if they are connected to the "determined edge pixel", otherwise they are also discarded. After the density raster image is processed through the above three steps, the Canny edge image is finally obtained. In the Canny edge image, the two edge pixel values of the curb are 255, and the other pixel values are 0.
  • stop circle points In the actual Canny edge image, not only the pixel value of the curb part is 255, but there are still circles of points with a pixel value of 255 in the image. For convenience of expression, these circle-by-circle points are referred to as “stop circle points" (as shown in FIG. 5A ).
  • stop circle points The reason for the "stop circle” is that when the collection vehicle is waiting for a red light, it needs to stop at the intersection.
  • the lidar has been working and collecting point clouds. Therefore, when the vehicle is parked, the number of points irradiated by the lidar will be more than that when the vehicle is not parked, and the density will be high.
  • the pixel value here in the density map is higher than the surrounding, so when extracting the Canny edge, this part will also be extracted as an edge. And this part of the pixels do not belong to the curb pixels, so they need to be removed.
  • the Canny edge map extracts both sides of the curb in the density map. Next, you need to distinguish the left and right edges of the curb.
  • the difference between the left and right edges of the curb is mainly reflected in the height, because in the real environment, the height of the curb is higher than the height of the road surface, and the height of the plane where the curb is located (such as the sidewalk, etc.) is higher than the height of the road surface The height value is high. Therefore, the height raster map can be used to filter the left and right edges of the curb.
  • the specific process is to traverse the points in the curb skeleton, assuming that there is a pixel point P k in the skeleton, in the 3*3 neighborhood of the P k pixel point, traverse the n of the 9 pixels whose pixel value is 255 in the Canny diagram points.
  • a 5*5 convolution is performed on the n pixels on the height grid map, and the convolution kernel is a 5*5 all-1 matrix.
  • the purpose of this is to select the curb edge points on the side near the road surface around P k .
  • the principle is that the surrounding height value of the curb pixel points on the side close to the road surface in the height map is smaller than that around the curb points on the side far away from the road surface. height value. Therefore, doing so can effectively distinguish the edges on both sides of the curb, and finally select the curb pixels close to the side of the road.
  • Step 405 output curb information.
  • the method for determining the curb provided by the above embodiments of the present disclosure can accurately determine the information of the curb by using the height grid map, the density grid map and the density grid map, so that the accuracy of fitting the curb is improved.
  • the present disclosure provides an embodiment of a curb determination device, which corresponds to the method embodiment shown in FIG. 2 , and the device specifically It can be applied to various electronic devices.
  • the curb determination device 600 of this embodiment includes: a point cloud frame acquisition unit 601 , a ground separation unit 602 , a grid division unit 603 , an information determination unit 604 and an information output unit 605 .
  • the point cloud frame acquisition unit 601 is configured to acquire point cloud frames collected at multiple collection points to obtain a sequence of point cloud frames.
  • the ground separation unit 602 is configured to determine the ground point cloud in each point cloud frame in the point cloud frame sequence.
  • the grid division unit 603 is configured to perform grid division after projecting the ground point cloud to the ground, and determine multiple grid images.
  • the information determining unit 604 is configured to determine road curb information according to multiple grid maps.
  • the information output unit 605 is configured to output curb information.
  • the ground separation unit 602 may be further configured to: for each point cloud frame in the point cloud frame sequence, perform morphological filtering on the point cloud frame to determine the point cloud The candidate ground point cloud in the frame; the point cloud points whose distance between the collection points corresponding to the point cloud frame in the candidate ground point cloud is greater than the preset distance threshold are deleted to obtain the ground point cloud.
  • the raster division unit 603 may be further configured to: divide the point cloud frame sequence according to the acquisition time corresponding to each point cloud frame in the point cloud frame sequence to obtain multiple Point cloud frame subsequence; splicing the ground point cloud in each point cloud frame in each point cloud frame subsequence; dividing the spliced ground point cloud to the ground and then performing grid division to obtain multiple grids; according to The intensity, density, and height of each point cloud point determine multiple raster maps.
  • the grid division unit 603 may be further configured to: determine the intensity, density, and height of each grid according to the intensity, density, and height of point cloud points included in each grid. Density and Height; Determines a map of multiple rasters based on the intensity, density, and height of each raster.
  • the plurality of raster maps include a density raster map, an intensity raster map, and a height raster map.
  • the information determining unit 604 may be further configured to: slice each density grid image, intensity grid image, and height grid image to obtain multiple density grid image slices, intensity grid image slices, and height grid image slices; According to the density raster image slice, intensity raster image slice and height raster image slice and the pre-trained segmentation model, determine the curb mask corresponding to each raster image slice; according to the density raster image, height raster image and each Curb mask, determine the curb information.
  • the information determining unit 604 may be further configured to: in the subsequence of point cloud frames corresponding to each raster image, according to the position of the collection point and the preset slice size, Multiple density raster images are sliced to obtain multiple density raster image slices.
  • the information determination unit 604 may be further configured to: splice each road tooth mask according to the relative position between the point cloud frame subsequences corresponding to each road tooth mask , to obtain the mosaic curb mask; according to the density raster map and the height raster map, post-process the mosaic curb mask to determine the curb information.
  • the information determining unit 604 may be further configured to: refine the spliced curb mask to determine candidate curb pixels; determine edge pixels according to the density grid map; Candidate curb pixels and edge pixels determine the curb edge pixels; determine the curb edge information according to the height grid map and curb edge pixels; fit the curb edge information to determine the curb information.
  • the units 601 to 605 described in the curb determining apparatus 600 respectively correspond to the steps in the method described with reference to FIG. 2 . Therefore, the operations and features described above for the method for determining the curb are also applicable to the device 600 and the units contained therein, and will not be repeated here.
  • the acquisition, storage and application of the user's personal information involved are in compliance with relevant laws and regulations, and do not violate public order and good customs.
  • the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.
  • FIG. 7 shows a block diagram of an electronic device 700 for executing a method for determining a road curb according to an embodiment of the present disclosure.
  • Electronic device is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers.
  • Electronic devices may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smart phones, wearable devices, and other similar computing devices.
  • the components shown herein, their connections and relationships, and their functions, are by way of example only, and are not intended to limit implementations of the disclosure described and/or claimed herein.
  • an electronic device 700 includes a processor 701 that can execute according to a computer program stored in a read-only memory (ROM) 702 or loaded from a memory 708 into a random access memory (RAM) 703. Various appropriate actions and treatments. In the RAM 703, various programs and data necessary for the operation of the electronic device 700 can also be stored.
  • the processor 701, ROM 702, and RAM 703 are connected to each other through a bus 704.
  • An I/O interface (input/output interface) 705 is also connected to the bus 704 .
  • the I/O interface 705 includes: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a memory 708, such as a magnetic disk, an optical disk, etc. ; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, and the like.
  • the communication unit 709 allows the electronic device 700 to exchange information/data with other devices through a computer network such as the Internet and/or various telecommunication networks.
  • Processor 701 may be various general and/or special purpose processing components having processing and computing capabilities. Some examples of processor 701 include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), various dedicated artificial intelligence (AI) computing chips, various processors that run machine learning model algorithms, digital signal processing processor (DSP), and any suitable processor, controller, microcontroller, etc.
  • the processor 701 executes various methods and processes described above, such as a method for determining curbs.
  • the curb determination method may be implemented as a computer software program tangibly embodied on a machine-readable storage medium, such as memory 708 .
  • part or all of the computer program can be loaded and/or installed on the electronic device 700 via the ROM 702 and/or the communication unit 709.
  • the computer program is loaded into the RAM 703 and executed by the processor 701, one or more steps of the above-described curb determination method can be performed.
  • the processor 701 may be configured in any other appropriate way (for example, by means of firmware) to execute the road curb determination method.
  • Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips Implemented in a system of systems (SOC), load programmable logic device (CPLD), computer hardware, firmware, software, and/or combinations thereof.
  • FPGAs field programmable gate arrays
  • ASICs application specific integrated circuits
  • ASSPs application specific standard products
  • SOC system of systems
  • CPLD load programmable logic device
  • computer hardware firmware, software, and/or combinations thereof.
  • programmable processor can be special-purpose or general-purpose programmable processor, can receive data and instruction from storage system, at least one input device, and at least one output device, and transmit data and instruction to this storage system, this at least one input device, and this at least one output device an output device.
  • Program codes for implementing the methods of the present disclosure may be written in any combination of one or more programming languages.
  • the above program code can be packaged into a computer program product.
  • These program codes or computer program products may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus, so that the program codes, when executed by the processor 701, make the flow diagrams and/or block diagrams specified The function/operation is implemented.
  • the program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.
  • a machine-readable storage medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device.
  • the machine-readable storage medium may be a machine-readable signal storage medium or a machine-readable storage medium.
  • a machine-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing.
  • machine-readable storage media would include one or more wire-based electrical connections, portable computer discs, hard drives, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
  • RAM random access memory
  • ROM read only memory
  • EPROM or flash memory erasable programmable read only memory
  • CD-ROM compact disk read-only memory
  • magnetic storage devices or any suitable combination of the foregoing.
  • the systems and techniques described herein can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user. ); and a keyboard and pointing device (eg, a mouse or a trackball) through which a user can provide input to the computer.
  • a display device e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor
  • a keyboard and pointing device eg, a mouse or a trackball
  • Other kinds of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and may be in any form (including Acoustic input, speech input or, tactile input) to receive input from the user.
  • the systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., as a a user computer having a graphical user interface or web browser through which a user can interact with embodiments of the systems and techniques described herein), or including such backend components, middleware components, Or any combination of front-end components in a computing system.
  • the components of the system can be interconnected by any form or medium of digital data communication, eg, a communication network. Examples of communication networks include: Local Area Network (LAN), Wide Area Network (WAN) and the Internet.
  • a computer system may include clients and servers.
  • Clients and servers are generally remote from each other and typically interact through a communication network.
  • the relationship of client and server arises by computer programs running on the respective computers and having a client-server relationship to each other.
  • the server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the problem of traditional physical host and VPS service ("Virtual Private Server", or "VPS”) Among them, there are defects such as difficult management and weak business scalability.
  • the server can also be a server of a distributed system, or a server combined with a blockchain.
  • steps may be reordered, added or deleted using the various forms of flow shown above.
  • each step described in the present disclosure may be executed in parallel, sequentially, or in a different order, as long as the desired result of the technical solution of the present disclosure can be achieved, no limitation is imposed herein.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Remote Sensing (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Electromagnetism (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

本公开提供了路牙确定方法、装置、设备以及存储介质，涉及数据处理领域。具体实现方案为：获取在多个采集点采集的点云帧，得到点云帧序列；确定点云帧序列中各点云帧中的地面点云；将地面点云向地面投影后进行栅格划分，确定多个栅格图；根据多个栅格图，确定路牙信息；输出路牙信息。本实现方式可以直接通过三维点云直接确定出路牙信息，提高了路牙信息的精度。

Description

路牙确定方法、装置、设备以及存储介质
本专利申请要求于2021年9月18日提交的、申请号为202111097024.X、发明名称为“路牙确定方法、装置、设备以及存储介质”的中国专利申请的优先权，该申请的全文以引用的方式并入本申请中。
技术领域
本公开涉及计算机技术领域，具体涉及数据处理领域，尤其涉及路牙确定方法、装置、设备以及存储介质。
背景技术
高精度地图是与普通电子导航地图不同的一种地图形式，高精度地图是服务于自动驾驶系统的地图。高精度地图由点云地图和矢量地图组成。点云地图是指由激光传感器采集到的环境地图。激光扫描周围环境，利用空间中的点云表示环境信息，最终将环境用三维点云的形式表示出来。因为激光的感受范围是360度，以及俯仰角的角度范围比较大，因此环境中的大部分信息均会被采集到，比如路面、车辆、行人、树木和建筑物等。矢量地图是指路面中包含的信息，比如路面上绘制的与交通规范相关的车道线或马路牙等等。
发明内容
本公开提供了一种路牙确定方法、装置、设备、存储介质以及计算机程序产品。
根据第一方面，提供了一种路牙确定方法，包括：获取在多个采集点采集的点云帧，得到点云帧序列；确定点云帧序列中各点云帧中的地面点云；将地面点云向地面投影后进行栅格划分，确定多个栅格图；根据多个栅格图，确定路牙信息；输出路牙信息。
根据第二方面，提供了一种路牙确定装置，包括：点云帧获取单元，被配置成获取在多个采集点采集的点云帧，得到点云帧序列；地面分离单元，被配置成确定点云帧序列中各点云帧中的地面点云；栅格划分单元，被配置成将地面点云向地面投影后进行栅格划分，确定多个栅格图；信息 确定单元，被配置成根据多个栅格图，确定路牙信息；信息输出单元，被配置成输出路牙信息。
根据第三方面，提供了一种电子设备，包括：至少一个处理器；以及与上述至少一个处理器通信连接的存储器；其中，存储器存储有可被至少一个处理器执行的指令，上述指令被至少一个处理器执行，以使至少一个处理器能够执行如第一方面所描述的方法。
根据第四方面，提供了一种存储有计算机指令的非瞬时计算机可读存储介质，上述计算机指令用于使计算机执行如第一方面所描述的方法。
根据第五方面，一种计算机程序产品，包括计算机程序，上述计算机程序在被处理器执行时实现如第一方面所描述的方法。
根据本公开的技术可以直接通过三维点云直接确定出路牙信息，提高了路牙信息的精度。
应当理解，本部分所描述的内容并非旨在标识本公开的实施例的关键或重要特征，也不用于限制本公开的范围。本公开的其它特征将通过以下的说明书而变得容易理解。
附图说明
附图用于更好地理解本方案，不构成对本公开的限定。其中：
图1是本公开的一个实施例可以应用于其中的示例性系统架构图；
图2是根据本公开的路牙确定方法的一个实施例的流程图；
图3是根据本公开的路牙确定方法的一个应用场景的示意图；
图4是根据本公开的路牙确定方法的另一个实施例的流程图；
图5A是图4所示实施例的“停止圈点”的示意图；
图5B是图4所示实施例的原始密度栅格图；
图5C是图4所示实施例的改进后的边缘算子的示意图；
图5D是图4所示实施例的强度栅格图；
图5E是图4所示实施例的canny边缘图；
图5F是图4所示实施例的高度栅格图；
图6是根据本公开的路牙确定装置的一个实施例的结构示意图；
图7是用来实现本公开实施例的路牙确定方法的电子设备的框图。
具体实施方式
以下结合附图对本公开的示范性实施例做出说明，其中包括本公开实施例的各种细节以助于理解，应当将它们认为仅仅是示范性的。因此，本领域普通技术人员应当认识到，可以对这里描述的实施例做出各种改变和修改，而不会背离本公开的范围和精神。同样，为了清楚和简明，以下的描述中省略了对公知功能和结构的描述。
需要说明的是，在不冲突的情况下，本公开中的实施例及实施例中的特征可以相互组合。下面将参考附图并结合实施例来详细说明本公开。
图1示出了可以应用本公开的路牙确定方法或路牙确定装置的实施例的示例性系统架构100。
如图1所示，系统架构100可以包括车辆101、网络102、服务器103和终端设备104。网络102用以在车辆101、服务器103和终端设备104之间提供通信链路的介质。网络102可以包括各种连接类型，例如有线、无线通信链路或者光纤电缆等等。
车辆101上可以安装有激光雷达，激光雷达用于采集车辆101周围环境的点云数据。上述点云数据可以包括多个点云帧。车辆101可以将采集的点云数据通过网络102发送给服务器103或者终端设备104。
服务器103可以对接收到的点云数据进行处理，以确定出路牙信息。并可以将路牙信息反馈给车辆101或者终端设备104。车辆101可以是自动驾驶车辆，可以根据高精地图和路牙信息更好的自动驾驶。
服务器103可以是硬件，也可以是软件。当服务器103为硬件时，可以实现成多个服务器组成的分布式服务器集群，也可以实现成单个服务器。当服务器103为软件时，可以实现成多个软件或软件模块(例如用来提供分布式服务)，也可以实现成单个软件或软件模块。在此不做具体限定。
终端设备104也可以接收点云数据，并对点云数据进行处理，并将确定出的路牙信息显示出来。终端设备104上可以安装有点云数据处理应用，可以对点云数据进行处理。用户可以通过终端设备104查看确定出的路牙信息。
需要说明的是，本公开实施例所提供的路牙确定方法可以由服务器103或终端设备104执行。相应地，路牙确定装置可以设置于服务器103或终端设备104中。
应该理解，图1中的车辆、网络、服务器和终端设备的数目仅仅是示意性的。根据实现需要，可以具有任意数目的车辆、网络、服务器和终端设备。
继续参考图2，其示出了根据本公开的路牙确定方法的一个实施例的流程200。本实施例的路牙确定方法，包括以下步骤：
步骤201，获取在多个采集点采集的点云帧，得到点云帧序列。
本实施例中，执行主体可以通过多种方式获取在多个采集点采集的点云帧。例如，执行主体可以获取采集车在行驶路径的多个采集点处采集的点云帧。各个采集点之间可以间隔预设距离。每个点云帧可以包括多个点云点，每个点云点可以包括高度信息和强度信息。不同采集点对应的点云帧可以形成点云帧序列。
步骤202，确定点云帧序列中各点云帧中的地面点云。
本实施例中，执行主体在得到点云帧序列后，可以对各点云帧进行地面分离，确定各点云帧中的地面点云。具体的中，执行主体可以将各点云帧中高度值小于预设高度阈值的点云点作为地面点云。或者，执行主体可以将各点云帧中除去已识别出的障碍物的点云点之后剩余的点云点作为地面点云。
步骤203，将地面点云向地面投影后进行栅格划分，确定多个栅格图。
在确定出各点云帧中的地面点云后，执行主体可以将各点云帧的地面点云向地面进行投影，然后对投影的各点云点进行栅格划分，得到多个栅格图。在栅格划分时可以以固定尺寸的栅格划分各投影点，这样，每个栅格中可以包括多个点云点的投影。执行主体可以根据各点云点的信息，确定每个栅格的信息。例如，可以根据点云点的强度、高度等信息，确定栅格的强度和高度信息。栅格图可以包括多个栅格，每个栅格的信息组成栅格图。栅格图可以包括高度栅格图、密度栅格图和强度栅格图。其中，高度栅格图是通过栅格内所有的点云点的高度值的中位数或平均数作为栅格的高度值。密度栅格图是通过栅格内所有的点云点的数量作为 栅格的密度值。强度栅格图是通过栅格内所有的点云点的强度值的中位数或平均数作为栅格的强度值。
步骤204，根据多个栅格图，确定路牙信息。
执行主体在确定多个栅格图后，可以通过多种方式确定路牙信息。例如，执行主体可以将各栅格图输入预先训练的路牙信息确定模型中，将模型的输出作为路牙信息。或者，执行主体可以对各栅格图进行图像处理，得到路牙信息。在进行图像处理时可以利用现有的图像处理算法，例如边缘检测算法、腐蚀算法、膨胀算法等。
步骤205，输出路牙信息。
执行主体在确定路牙信息后，可以输出上述路牙信息。具体的，执行主体可以将路牙信息输出给自动驾驶车辆，以供自动驾驶车辆根据路牙信息进行停车等操作。或者，执行主体可以将路牙信息输出给终端的屏幕，以供用户查看。
继续参见图3，其示出了根据本公开的路牙确定方法的一个应用场景的示意图。在图3的应用场景中，采集车301在行驶路径的各个采集点处采集周围环境的点云数据。然后将才采集的点云数据发送给服务器302。服务器302可以对点云数据进行处理，确定出周围环境的路牙信息。然后将路牙信息更新到高精地图中，并将更新后的高精地图发送给自动驾驶车辆303，自动驾驶车辆303可以根据更新后的高精地图更准确地停车。
本公开的上述实施例提供的路牙确定方法，可以通过对点云帧进行栅格划分，利用栅格图确定路牙信息，从而提高了路牙确定的精度。
继续参见图4，其示出了根据本公开的路牙确定方法的另一个实施例的流程400。如图4所示，本实施例的方法可以包括以下步骤：
步骤401，获取在多个采集点采集的点云帧，得到点云帧序列。
步骤402，对于点云帧序列中的每个点云帧，对该点云帧进行形态学滤波，确定该点云帧中的候选地面点云；将候选地面点云中与该点云帧对应的采集点之间的距离大于预设距离阈值的点云点删除，得到地面点云。
本实施例中，执行主体可以对点云帧序列中的每个点云帧进行处理，即对每个点云帧进行形态学滤波，实现地面分离。将滤波后得到的点云作为候选地面点云。这里，形态学滤波可以包括多种处理，例如膨胀、腐蚀 等处理。然后，执行主体可以将候选地面点云中与该点云帧对应的采集点之间的距离大于预设距离阈值的点云点删除，得到地面点云。由于马路牙一般位于地面点云里，所以这里只对地面点云进行处理，可以加快处理速度。这里，距离采集点较远的点云点代表距离车体较远的点云点，此处的点云稀疏并且可信度较差，所以这里将不可信点云点删除。
步骤403，根据点云帧序列中各点云帧对应的采集时间，对点云帧序列进行划分，得到多个点云帧子序列；对各点云帧子序列中的各点云帧中的地面点云进行拼接；将拼接后的地面点云向地面投影后进行栅格划分，得到多个栅格；根据各点云点的强度、密度和高度，确定多个栅格图。
本实施例中，执行主体可以首先对点云帧序列中的各点云帧按照时间进行划分。即将各点云帧对应的采集时间按每隔30S对点云帧序列进行划分，得到多个点云帧子序列。这里，考虑到点云数据的体量较大，单次对上述点云帧序列进行处理，可能导致处理速度变慢。所以，这里将点云帧序列进行划分后再进行处理，可以保证点云处理的平稳进行。执行主体可以分别对各点云帧子序列中的各点云帧中的地面点云进行拼接。在拼接时，可以根据GPS和惯性导航单据估计的位姿进行拼接。然后，执行主体可以将拼接后的点云向地面进行投影后进行栅格划分，得到多个栅格。然后，根据每个栅格中包括的点云点的强度、密度和高度，确定多个栅格图。这里，可以设置每个栅格为一个正方形，正方形的边长可以根据实际情况进行设定。
在本实施例的一些可选的实现方式中，执行主体可以通过以下步骤确定多个栅格图：根据每个栅格中包括的点云点的强度、密度和高度，确定每个栅格的强度、密度和高度；根据每个栅格的强度、密度和高度，确定多个栅格图。
本实现方式中，执行主体可以根据每个栅格中包括的点云点的强度、密度和高度，确定每个栅格的强度、密度和高度。然后根据每个栅格的强度、密度和高度，确定多个栅格图。具体的，强度栅格图是通过栅格内所有的点云点的强度值的中位数或平均数作为栅格的强度值。高度栅格图是通过栅格内所有的点云点的高度值的中位数或平均数作为栅格的高度值。 密度栅格图是通过栅格内所有的点云的数量作为栅格的密度值。
步骤404，对各密度栅格图、强度栅格图和高度栅格图进行切片，得到多个密度栅格图切片、强度栅格图切片和高度栅格图切片；根据密度栅格图切片、强度栅格图切片和高度栅格图切片以及预先训练的分割模型，确定各栅格图切片对应的路牙掩码；根据密度栅格图、高度栅格图以及各路牙掩码，确定路牙信息。
通过以上三种栅格图基本可以反应环境的地面信息。但是将点云数据按照时间划分存在一个缺点，比如在栅格的边缘部分，因为没有下一个时间点的点云数据，因此也没有对应的栅格像素，所以会导致栅格的边缘可能存在像素点稀疏的情况。这样对后续的处理也会造成负面影响。因此本文对时间划分的栅格进行重组，将其按照空间划分，这样按照空间划分后的栅格就不存在上述的缺点。接下来通过深度学习模型分割出栅格中的马路牙，但是按照空间划分的栅格图尺寸很大，如果直接读取会将内存爆满，无法直接放入到模型中。
本实施例中，执行主体可以按空间对多个栅格图进行切片，得到多个栅格图切片。切片的尺寸恰好可以送入深度学习模型中检测。然后，执行主体可以将多个栅格图切片输入预先训练的分割模型，确定对应的路牙掩码。具体的，执行主体可以分别对各密度栅格图、强度栅格图以及强度栅格图进行切片，然后将得到的各密度栅格图切片、强度栅格图切片以及强度栅格图切片组成三通道图片输入预先训练的分割模型，确定对应的路牙掩码。可以理解的是，输入模型的密度栅格图切片、强度栅格图切片以及强度栅格图切片来自于同一点云帧子序列。同样的，得到的每个路牙掩码与单个点云帧子序列对应。
执行主体可以将各路牙掩码进行拼接，得到整条道路的路牙掩码。执行主体可以直接将路牙掩码作为路牙信息。或者，执行主体还可以进一步对路牙掩码进行进一步处理，以得到更清晰更准确的路牙信息。
在本实施例的一些可选的实现方式中，执行主体可以通过以下步骤确定栅格图切片：在各栅格图对应的点云帧子序列中，根据采集点的位置以及预设的切片尺寸，对所述多个密度栅格图进行切片，得到多个密度栅格 图切片。
实现方式中，执行主体可以根据采集点的位置以及预设的切片尺寸，对所述多个密度栅格图进行切片，得到多个密度栅格图切片。具体的，执行主体可以等间隔地选取采集点，在采集点周围框定一个空间范围，切成960*640像素，切片大小恰好可以送入模型中检测。本公开选取的是基于Deeplabv3框架研发的模型。
在本实施例的一些可选的实现方式中，执行主体可以通过以下步骤确定路牙信息：根据各路牙掩码对应的点云帧子序列之间的相对位置，对各路牙掩码进行拼接，得到拼接路牙掩码；根据密度栅格图、高度栅格图，对拼接路牙掩码进行后处理，确定路牙信息。
本实现方式中，执行主体可以根据各路牙掩码对应的点云帧子序列之间的相对位置，对各路牙掩码进行拼接，得到拼接路牙掩码。即将相邻的点云帧子序列对应的路牙掩码进行拼接。在将各路牙掩码拼接后可以得到拼接路牙掩码。执行主体可以进一步结合密度栅格图、高度栅格图，对拼接路牙掩码进行拟合等后处理，确定准确的路牙信息。
在本实施例的一些可选的实现方式中，执行主体可以通过以下步骤进一步对拼接路牙掩码进行后处理：对拼接路牙掩码进行细化，确定候选路牙像素；根据密度栅格图，确定边缘像素；根据候选路牙像素以及边缘像素，确定路牙边缘像素；根据高度栅格图以及路牙边缘像素，确定路牙的边缘信息；对路牙的边缘信息进行拟合，确定路牙信息。
本实现方式中，执行主体可以首先对拼接路牙掩码进行细化，目的是将每一条马路牙腐蚀成为没有宽度的一条线。细化后的掩码图已经基本能看出马路牙的轮廓，只不过细化算法后形状不规整，不是一条平滑的直线，不能够直接拟合，所以需要后处理操作。这里，称细化后的像素为候选路牙像素。
后处理的目标是将拼接路牙掩码中的候选路牙像素拟合为几何信息，但是候选路牙像素是有宽度的，因为实际中的马路牙也是有宽度的。而目标需要输出的马路牙矢量是没有宽度。上述深度学习模型输出的路牙掩码是在靠近路面一侧的马路牙边缘。本实施例是为了确认马路牙的有效轮廓， 保证自动驾驶车辆的安全通行，在拟合之前需要对拼接路牙掩码图中的候选路牙像素进行筛选，筛选掉远离路面一侧的马路牙像素点。这样才能提升后续的马路牙拟合精度。
筛选像素点需要密度栅格图和高度栅格图两个栅格图作为参考。因为马路牙存在高度，所以激光雷达发射的激光照射在马路牙上的点云数量会大于普通地面点。因此在密度栅格图中，马路牙部分的像素值也会比普通地面部分高，因此在密度栅格图中，马路牙的像素点属于边缘点。这里，首先对密度栅格图提取边缘信息，称为边缘像素。这里，采用Canny边缘检测算子作用在密度栅格图上。Canny边缘算法是在Sobel边缘算法的基础上，对边缘像素进行更细致的后处理，过滤边缘点周围的边缘点，从而使得边缘部分更加细致准确。Canny边缘检测可以细分为三步：1.采用Sobel卷积核进行卷积运算。Sobel在x方向的卷积核如公式1所示，Sobel在y方向的卷积核公式2式所示。
Figure PCTCN2022104941-appb-000001
Figure PCTCN2022104941-appb-000002
然后基于得到x方向和y方向的梯度值的G x和G y梯度值计算整体梯度值
Figure PCTCN2022104941-appb-000003
方向角
Figure PCTCN2022104941-appb-000004
基于边缘梯度方向采用非极大值抑制，目的是选取一个范围内梯度值最大的像素点，目的是使提取到的边缘更加稳定，避免提取到密度栅格图中的噪点。
设定两个阈值T 1，T 2。梯度值大于T 2的像素点被归为“确定边缘像素”被保留。梯度值小与T 1的像素点被认为一定不属于边缘，被丢弃。梯度值介于T 1和T 2之间的像素点，如果它们连接到“确定边缘像素”，则它们被视为边缘的一部分，否则也会被丢弃。密度栅格图经过了以上三个步骤处理后最终得到Canny边缘图像。Canny边缘图像中马路牙的两个边缘像素值为255，其他像素值为0。
实际Canny边缘图像中不仅仅只有马路牙部分的像素值是255，图像 中仍然存在一圈一圈像素值为255的点。为了表达方便将这些一圈一圈的点称之为“停止圈点”(如图5A所示)。造成出现“停止圈点”的原因是，采集车辆在等红灯时，需要在路口停车。在停车的过程中，激光雷达一直工作，一直采集点云。所以在车辆停车时，激光雷达照射到的地方的点数量就会比不停车时多，密度大。因此在密度图中此处的像素值与周围相比要高，因此提取Canny边缘时，这部分也会被当作边缘提取出来。而这一部分的像素点并不属于马路牙像素点，因此需要被剔除。在密度栅格图中可以发现，一圈圈点的像素值要大于马路牙的像素值，原因是“停止圈点”在点云地图中的密度值要更大。因此设立阈值T lut＝90用于区分“停止圈点”和马路牙点。首先设立一个查找表，查找表的自变量为x，因变量设为y。两者的映射关系如公式3所示。
Figure PCTCN2022104941-appb-000005
首先将查找表算法应用到原始的密度栅格图I 1上，然后进行一次中值滤波，把“停止圈点”周围的像素也进行模糊处理得到I filter。之后把原始密度栅格图I 1(如图5B所示)中像素值大于T lut＝90的像素点替换成I filter中相同位置的像素值，最终得到处理后的I canny(如图5C所示)。
Canny边缘图将密度图中的马路牙的两侧提取出来了。接下来需要区分马路牙的左右两侧边缘。马路牙的左右两个边缘的不同点主要体现在高度上，因为现实环境中，马路牙的高度要高于路面高度，同时马路牙所在的平面(比如人行道等)的高度值，均比路面的高度值高。因此高度栅格图可以用来筛选马路牙的左右两个边缘。具体的流程是遍历马路牙骨架中的点，假设骨架中有一个像素点P k，在P k像素点的3*3邻域内，遍历这9个像素中在Canny图中像素值为255的n个点。为了保证稳定，对这n个像素点在高度栅格图上进行5*5的卷积，卷积核是5*5的全1矩阵。这样做的目的是为了选取P k周围靠近路面一侧的马路牙边缘点，原理是靠近路面一侧的马路牙像素点在高度图中的周围高度值要小于远离路面一侧的马路牙点周围的高度值。因此这样做可以有效区分出马路牙两侧的边缘，并最终选取到靠近路面一侧的马路牙像素点。
取得马路牙候选像素点之后，可以对马路牙候选像素点进行分段拟合。 每一段拟合采用道格拉斯普克算法，目的是使拟合的每段马路牙都是折线的形式。最终拟合出的马路牙与路面实际的马路牙相重合。如图5D-5f分别表示强度栅格图、canny边缘图和高度栅格图。图中的直线表示拟合后的马路牙矢量。
步骤405，输出路牙信息。
本公开的上述实施例提供的路牙确定方法，可以利用高度栅格图、密度栅格图和密度栅格图准确地确定出路牙信息，使到拟合马路牙精度得到提升。
进一步参考图6，作为对上述各图所示方法的实现，本公开提供了一种路牙确定装置的一个实施例，该装置实施例与图2所示的方法实施例相对应，该装置具体可以应用于各种电子设备中。
如图6所示，本实施例的路牙确定装置600包括：点云帧获取单元601、地面分离单元602、栅格划分单元603、信息确定单元604和信息输出单元605。
点云帧获取单元601，被配置成获取在多个采集点采集的点云帧，得到点云帧序列。
地面分离单元602，被配置成确定点云帧序列中各点云帧中的地面点云。
栅格划分单元603，被配置成将地面点云向地面投影后进行栅格划分，确定多个栅格图。
信息确定单元604，被配置成根据多个栅格图，确定路牙信息。
信息输出单元605，被配置成输出路牙信息。
在本实施例的一些可选的实现方式中，地面分离单元602可以进一步被配置成：对于点云帧序列中的每个点云帧，对该点云帧进行形态学滤波，确定该点云帧中的候选地面点云；将候选地面点云中与该点云帧对应的采集点之间的距离大于预设距离阈值的点云点删除，得到地面点云。
在本实施例的一些可选的实现方式中，栅格划分单元603可以进一步被配置成：根据点云帧序列中各点云帧对应的采集时间，对点云帧序列进行划分，得到多个点云帧子序列；对各点云帧子序列中的各点云帧中的地 面点云进行拼接；将拼接后的地面点云向地面投影后进行栅格划分，得到多个栅格；根据各点云点的强度、密度和高度，确定多个栅格图。
在本实施例的一些可选的实现方式中，栅格划分单元603可以进一步被配置成：根据每个栅格中包括的点云点的强度、密度和高度，确定每个栅格的强度、密度和高度；根据每个栅格的强度、密度和高度，确定多个栅格图。
在本实施例的一些可选的实现方式中，多个栅格图包括密度栅格图、强度栅格图和高度栅格图。信息确定单元604可以进一步被配置成：对各密度栅格图、强度栅格图和高度栅格图进行切片，得到多个密度栅格图切片、强度栅格图切片和高度栅格图切片；根据密度栅格图切片、强度栅格图切片和高度栅格图切片以及预先训练的分割模型，确定各栅格图切片对应的路牙掩码；根据密度栅格图、高度栅格图以及各路牙掩码，确定路牙信息。
在本实施例的一些可选的实现方式中，信息确定单元604可以进一步被配置成：在各栅格图对应的点云帧子序列中，根据采集点的位置以及预设的切片尺寸，对多个密度栅格图进行切片，得到多个密度栅格图切片。
在本实施例的一些可选的实现方式中，信息确定单元604可以进一步被配置成：根据各路牙掩码对应的点云帧子序列之间的相对位置，对各路牙掩码进行拼接，得到拼接路牙掩码；根据密度栅格图、高度栅格图，对拼接路牙掩码进行后处理，确定路牙信息。
在本实施例的一些可选的实现方式中，信息确定单元604可以进一步被配置成：对拼接路牙掩码进行细化，确定候选路牙像素；根据密度栅格图，确定边缘像素；对候选路牙像素以及边缘像素，确定路牙边缘像素；根据高度栅格图以及路牙边缘像素，确定路牙的边缘信息；对路牙的边缘信息进行拟合，确定路牙信息。
应当理解，路牙确定装置600中记载的单元601至单元605分别与参考图2中描述的方法中的各个步骤相对应。由此，上文针对路牙确定方法描述的操作和特征同样适用于装置600及其中包含的单元，在此不再赘述。
本公开的技术方案中，所涉及的用户个人信息的获取、存储和应用等， 均符合相关法律法规的规定，且不违背公序良俗。
根据本公开的实施例，本公开还提供了还提供了一种电子设备、一种可读存储介质和一种计算机程序产品。
图7示出了根据本公开实施例的执行路牙确定方法的电子设备700的框图。电子设备旨在表示各种形式的数字计算机，诸如，膝上型计算机、台式计算机、工作台、个人数字助理、服务器、刀片式服务器、大型计算机、和其它适合的计算机。电子设备还可以表示各种形式的移动装置，诸如，个人数字处理、蜂窝电话、智能电话、可穿戴设备和其它类似的计算装置。本文所示的部件、它们的连接和关系、以及它们的功能仅仅作为示例，并且不意在限制本文中描述的和/或者要求的本公开的实现。
如图7所示，电子设备700包括处理器701，其可以根据存储在只读存储器(ROM)702中的计算机程序或者从存储器708加载到随机访问存储器(RAM)703中的计算机程序，来执行各种适当的动作和处理。在RAM 703中，还可存储电子设备700操作所需的各种程序和数据。处理器701、ROM 702以及RAM 703通过总线704彼此相连。I/O接口(输入/输出接口)705也连接至总线704。
电子设备700中的多个部件连接至I/O接口705，包括：输入单元706，例如键盘、鼠标等；输出单元707，例如各种类型的显示器、扬声器等；存储器708，例如磁盘、光盘等；以及通信单元709，例如网卡、调制解调器、无线通信收发机等。通信单元709允许电子设备700通过诸如因特网的计算机网络和/或各种电信网络与其他设备交换信息/数据。
处理器701可以是各种具有处理和计算能力的通用和/或专用处理组件。处理器701的一些示例包括但不限于中央处理单元(CPU)、图形处理单元(GPU)、各种专用的人工智能(AI)计算芯片、各种运行机器学习模型算法的处理器、数字信号处理器(DSP)、以及任何适当的处理器、控制器、微控制器等。处理器701执行上文所描述的各个方法和处理，例如路牙确定方法。例如，在一些实施例中，路牙确定方法可被实现为计算机软件程序，其被有形地包含于机器可读存储介质，例如存储器708。在一些实施例中，计算机程序的部分或者全部可以经由ROM 702和/或通信单元709而被载入和/或安装到电子设备700上。当计算机程序加载到RAM  703并由处理器701执行时，可以执行上文描述的路牙确定方法的一个或多个步骤。备选地，在其他实施例中，处理器701可以通过其他任何适当的方式(例如，借助于固件)而被配置为执行路牙确定方法。
本文中以上描述的系统和技术的各种实施方式可以在数字电子电路系统、集成电路系统、场可编程门阵列(FPGA)、专用集成电路(ASIC)、专用标准产品(ASSP)、芯片上系统的系统(SOC)、负载可编程逻辑设备(CPLD)、计算机硬件、固件、软件、和/或它们的组合中实现。这些各种实施方式可以包括：实施在一个或者多个计算机程序中，该一个或者多个计算机程序可在包括至少一个可编程处理器的可编程系统上执行和/或解释，该可编程处理器可以是专用或者通用可编程处理器，可以从存储系统、至少一个输入装置、和至少一个输出装置接收数据和指令，并且将数据和指令传输至该存储系统、该至少一个输入装置、和该至少一个输出装置。
用于实施本公开的方法的程序代码可以采用一个或多个编程语言的任何组合来编写。上述程序代码可以封装成计算机程序产品。这些程序代码或计算机程序产品可以提供给通用计算机、专用计算机或其他可编程数据处理装置的处理器或控制器，使得程序代码当由处理器701执行时使流程图和/或框图中所规定的功能/操作被实施。程序代码可以完全在机器上执行、部分地在机器上执行，作为独立软件包部分地在机器上执行且部分地在远程机器上执行或完全在远程机器或服务器上执行。
在本公开的上下文中，机器可读存储介质可以是有形的介质，其可以包含或存储以供指令执行系统、装置或设备使用或与指令执行系统、装置或设备结合地使用的程序。机器可读存储介质可以是机器可读信号存储介质或机器可读存储介质。机器可读存储介质可以包括但不限于电子的、磁性的、光学的、电磁的、红外的、或半导体系统、装置或设备，或者上述内容的任何合适组合。机器可读存储介质的更具体示例会包括基于一个或多个线的电气连接、便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦除可编程只读存储器(EPROM或快闪存储器)、光纤、便捷式紧凑盘只读存储器(CD-ROM)、光学存储设备、磁存储设备、或上述内容的任何合适组合。
为了提供与用户的交互，可以在计算机上实施此处描述的系统和技术，该计算机具有：用于向用户显示信息的显示装置(例如，CRT(阴极射线管)或者LCD(液晶显示器)监视器)；以及键盘和指向装置(例如，鼠标或者轨迹球)，用户可以通过该键盘和该指向装置来将输入提供给计算机。其它种类的装置还可以用于提供与用户的交互；例如，提供给用户的反馈可以是任何形式的传感反馈(例如，视觉反馈、听觉反馈、或者触觉反馈)；并且可以用任何形式(包括声输入、语音输入或者、触觉输入)来接收来自用户的输入。
可以将此处描述的系统和技术实施在包括后台部件的计算系统(例如，作为数据服务器)、或者包括中间件部件的计算系统(例如，应用服务器)、或者包括前端部件的计算系统(例如，具有图形用户界面或者网络浏览器的用户计算机，用户可以通过该图形用户界面或者该网络浏览器来与此处描述的系统和技术的实施方式交互)、或者包括这种后台部件、中间件部件、或者前端部件的任何组合的计算系统中。可以通过任何形式或者介质的数字数据通信(例如，通信网络)来将系统的部件相互连接。通信网络的示例包括：局域网(LAN)、广域网(WAN)和互联网。
计算机系统可以包括客户端和服务器。客户端和服务器一般远离彼此并且通常通过通信网络进行交互。通过在相应的计算机上运行并且彼此具有客户端-服务器关系的计算机程序来产生客户端和服务器的关系。服务器可以是云服务器，又称为云计算服务器或云主机，是云计算服务体系中的一项主机产品，以解决了传统物理主机与VPS服务(“Virtual Private Server”，或简称“VPS”)中，存在的管理难度大，业务扩展性弱的缺陷。服务器也可以是分布式系统的服务器，或者是结合了区块链的服务器。
应该理解，可以使用上面所示的各种形式的流程，重新排序、增加或删除步骤。例如，本公开中记载的各步骤可以并行地执行也可以顺序地执行也可以不同的次序执行，只要能够实现本公开的技术方案所期望的结果，本文在此不进行限制。
上述具体实施方式，并不构成对本公开保护范围的限制。本领域技术人员应该明白的是，根据设计要求和其他因素，可以进行各种修改、组合、子组合和替代。任何在本公开的精神和原则之内所作的修改、等同替换和 改进等，均应包含在本公开的保护范围之内。

Claims (19)

  1. 一种路牙确定方法，包括：
    获取在多个采集点采集的点云帧，得到点云帧序列；
    确定所述点云帧序列中各点云帧中的地面点云；
    将所述地面点云向地面投影后进行栅格划分，确定多个栅格图；
    根据所述多个栅格图，确定路牙信息；
    输出所述路牙信息。
  2. 根据权利要求1所述的方法，其中，所述确定所述点云帧序列中各点云帧中的地面点云，包括：
    对于所述点云帧序列中的每个点云帧，对该点云帧进行形态学滤波，确定该点云帧中的候选地面点云；
    将所述候选地面点云中与该点云帧对应的采集点之间的距离大于预设距离阈值的点云点删除，得到地面点云。
  3. 根据权利要求1所述的方法，其中，所述将所述地面点云向地面投影后进行栅格划分，确定多个栅格图，包括：
    根据所述点云帧序列中各点云帧对应的采集时间，对所述点云帧序列进行划分，得到多个点云帧子序列；
    对各点云帧子序列中的各点云帧中的地面点云进行拼接；
    将拼接后的地面点云向地面投影后进行栅格划分，得到多个栅格；
    根据各点云点的强度、密度和高度，确定多个栅格图。
  4. 根据权利要求3所述的方法，其中，所述根据各点云点的强度、密度和高度，确定多个栅格图，包括：
    根据每个栅格中包括的点云点的强度、密度和高度，确定每个栅格的强度、密度和高度；
    根据每个栅格的强度、密度和高度，确定多个栅格图。
  5. 根据权利要求1所述的方法，其中，所述多个栅格图包括密度栅格图、强度栅格图和高度栅格图；以及
    所述根据所述多个栅格图，确定路牙信息，包括：
    对各密度栅格图、强度栅格图和高度栅格图进行切片，得到多个密度栅格图切片、强度栅格图切片和高度栅格图切片；
    根据所述密度栅格图切片、强度栅格图切片和高度栅格图切片以及预先训练的分割模型，确定各栅格图切片对应的路牙掩码；
    根据所述密度栅格图、所述高度栅格图以及各所述路牙掩码，确定路牙信息。
  6. 根据权利要求5所述的方法，其中，所述对所述多个栅格图进行切片，得到多个栅格图切片，包括：
    在各栅格图对应的点云帧子序列中，根据采集点的位置以及预设的切片尺寸，对所述多个密度栅格图进行切片，得到多个密度栅格图切片。
  7. 根据权利要求5所述的方法，其中，所述根据所述密度栅格图、所述高度栅格图以及各所述路牙掩码，确定路牙信息，包括：
    根据各路牙掩码对应的点云帧子序列之间的相对位置，对各所述路牙掩码进行拼接，得到拼接路牙掩码；
    根据所述密度栅格图、所述高度栅格图，对所述拼接路牙掩码进行后处理，确定路牙信息。
  8. 根据权利要求7所述的方法，其中，所述根据所述密度栅格图、所述高度栅格图，对所述拼接路牙掩码进行后处理，确定路牙信息，包括：
    对所述拼接路牙掩码进行细化，确定候选路牙像素；
    根据所述密度栅格图，确定边缘像素；
    根据所述候选路牙像素以及所述边缘像素，确定路牙边缘像素；
    根据所述高度栅格图以及所述路牙边缘像素，确定路牙的边缘信息；
    对所述路牙的边缘信息进行拟合，确定路牙信息。
  9. 一种路牙确定装置，包括：
    点云帧获取单元，被配置成获取在多个采集点采集的点云帧，得到点云帧序列；
    地面分离单元，被配置成确定所述点云帧序列中各点云帧中的地面点云；
    栅格划分单元，被配置成将所述地面点云向地面投影后进行栅格划分，确定多个栅格图；
    信息确定单元，被配置成根据所述多个栅格图，确定路牙信息；
    信息输出单元，被配置成输出所述路牙信息。
  10. 根据权利要求9所述的装置，其中，所述地面分离单元进一步被配置成：
    对于所述点云帧序列中的每个点云帧，对该点云帧进行形态学滤波，确定该点云帧中的候选地面点云；
    将所述候选地面点云中与该点云帧对应的采集点之间的距离大于预设距离阈值的点云点删除，得到地面点云。
  11. 根据权利要求9所述的装置，其中，所述栅格划分单元进一步被配置成：
    根据所述点云帧序列中各点云帧对应的采集时间，对所述点云帧序列进行划分，得到多个点云帧子序列；
    对各点云帧子序列中的各点云帧中的地面点云进行拼接；
    将拼接后的地面点云向地面投影后进行栅格划分，得到多个栅格；
    根据各点云点的强度、密度和高度，确定多个栅格图。
  12. 根据权利要求11所述的装置，其中，所述栅格划分单元进一步被配置成：
    根据每个栅格中包括的点云点的强度、密度和高度，确定每个栅格的强度、密度和高度；
    根据每个栅格的强度、密度和高度，确定多个栅格图。
  13. 根据权利要求11所述的装置，其中，所述多个栅格图包括密度栅格图、强度栅格图和高度栅格图；以及
    所述信息确定单元进一步被配置成：
    对各密度栅格图、强度栅格图和高度栅格图进行切片，得到多个密度栅格图切片、强度栅格图切片和高度栅格图切片；
    根据所述密度栅格图切片、强度栅格图切片和高度栅格图切片以及预先训练的分割模型，确定各栅格图切片对应的路牙掩码；
    根据所述密度栅格图、所述高度栅格图以及各所述路牙掩码，确定路牙信息。
  14. 根据权利要求13所述的装置，其中，所述信息确定单元进一步被配置成：
    在各栅格图对应的点云帧子序列中，根据采集点的位置以及预设的切片尺寸，对所述多个密度栅格图进行切片，得到多个密度栅格图切片。
  15. 根据权利要求13所述的装置，其中，所述信息确定单元进一步被配置成：
    根据各路牙掩码对应的点云帧子序列之间的相对位置，对各所述路牙掩码进行拼接，得到拼接路牙掩码；
    根据所述密度栅格图、所述高度栅格图，对所述拼接路牙掩码进行后处理，确定路牙信息。
  16. 根据权利要求15所述的装置，其中，所述信息确定单元进 一步被配置成：
    对所述拼接路牙掩码进行细化，确定候选路牙像素；
    根据所述密度栅格图，确定边缘像素；
    根据所述候选路牙像素以及所述边缘像素，确定路牙边缘像素；
    根据所述高度栅格图以及所述路牙边缘像素，确定路牙的边缘信息；
    对所述路牙的边缘信息进行拟合，确定路牙信息。
  17. 一种电子设备，包括：
    至少一个处理器；以及
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行权利要求1-8中任一项所述的方法。
  18. 一种存储有计算机指令的非瞬时计算机可读存储介质，其中，所述计算机指令用于使所述计算机执行权利要求1-8中任一项所述的方法。
  19. 一种计算机程序产品，包括计算机程序，其中，所述计算机程序在被处理器执行时实现根据权利要求1-8中任一项所述的方法。
PCT/CN2022/104941 2021-09-18 2022-07-11 路牙确定方法、装置、设备以及存储介质 WO2023040437A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111097024.XA CN113806464A (zh) 2021-09-18 2021-09-18 路牙确定方法、装置、设备以及存储介质
CN202111097024.X 2021-09-18

Publications (1)

Publication Number Publication Date
WO2023040437A1 true WO2023040437A1 (zh) 2023-03-23

Family

ID=78895947

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/104941 WO2023040437A1 (zh) 2021-09-18 2022-07-11 路牙确定方法、装置、设备以及存储介质

Country Status (2)

Country Link
CN (1) CN113806464A (zh)
WO (1) WO2023040437A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116503434A (zh) * 2023-05-15 2023-07-28 中国电建集团昆明勘测设计研究院有限公司 一种点云数据的边界提取方法、装置、设备及存储介质
CN116990832A (zh) * 2023-08-04 2023-11-03 长沙行深智能科技有限公司 危险道路边沿感知方法、装置、终端设备及存储介质

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113806464A (zh) * 2021-09-18 2021-12-17 北京京东乾石科技有限公司 路牙确定方法、装置、设备以及存储介质
CN115131525B (zh) * 2022-07-26 2024-04-05 白犀牛智达(北京)科技有限公司 一种路牙检测方法

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150323672A1 (en) * 2012-11-22 2015-11-12 Geosim Systems Ltd. Point-cloud fusion
CN106908775A (zh) * 2017-03-08 2017-06-30 同济大学 一种基于激光反射强度的无人车实时定位方法
CN110942024A (zh) * 2019-11-26 2020-03-31 江苏天策机器人科技有限公司 一种基于深度学习的无人车路牙检测方法
CN111353969A (zh) * 2018-12-20 2020-06-30 长沙智能驾驶研究院有限公司 道路可行驶区域的确定方法、装置及计算机设备
CN112907747A (zh) * 2021-03-26 2021-06-04 上海商汤临港智能科技有限公司 一种点云数据处理的方法、装置、电子设备及存储介质
CN112988922A (zh) * 2019-12-16 2021-06-18 长沙智能驾驶研究院有限公司 感知地图构建方法、装置、计算机设备和存储介质
CN113281770A (zh) * 2021-05-28 2021-08-20 东软睿驰汽车技术（沈阳）有限公司 坐标系关系获取方法及装置
CN113806464A (zh) * 2021-09-18 2021-12-17 北京京东乾石科技有限公司 路牙确定方法、装置、设备以及存储介质

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109766878B (zh) * 2019-04-11 2019-06-28 深兰人工智能芯片研究院（江苏）有限公司 一种车道线检测的方法和设备
CN111999720B (zh) * 2020-07-08 2022-08-16 深圳市速腾聚创科技有限公司 激光雷达参数调整方法、激光雷达系统和计算机存储介质
CN112131947A (zh) * 2020-08-21 2020-12-25 河北鼎联科技有限公司 道路指示线提取方法及装置
CN112149572A (zh) * 2020-09-24 2020-12-29 知行汽车科技（苏州）有限公司 路沿检测方法、装置和存储介质
CN112347999B (zh) * 2021-01-07 2021-05-14 深圳市速腾聚创科技有限公司 障碍物识别模型训练方法、障碍物识别方法、装置及系统
CN112964264B (zh) * 2021-02-07 2024-03-26 上海商汤临港智能科技有限公司 道路边沿检测方法、装置、高精度地图、车辆及存储介质
CN113139982B (zh) * 2021-04-21 2023-03-14 香港理工大学深圳研究院 一种室内房间点云的自动分割方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150323672A1 (en) * 2012-11-22 2015-11-12 Geosim Systems Ltd. Point-cloud fusion
CN106908775A (zh) * 2017-03-08 2017-06-30 同济大学 一种基于激光反射强度的无人车实时定位方法
CN111353969A (zh) * 2018-12-20 2020-06-30 长沙智能驾驶研究院有限公司 道路可行驶区域的确定方法、装置及计算机设备
CN110942024A (zh) * 2019-11-26 2020-03-31 江苏天策机器人科技有限公司 一种基于深度学习的无人车路牙检测方法
CN112988922A (zh) * 2019-12-16 2021-06-18 长沙智能驾驶研究院有限公司 感知地图构建方法、装置、计算机设备和存储介质
CN112907747A (zh) * 2021-03-26 2021-06-04 上海商汤临港智能科技有限公司 一种点云数据处理的方法、装置、电子设备及存储介质
CN113281770A (zh) * 2021-05-28 2021-08-20 东软睿驰汽车技术（沈阳）有限公司 坐标系关系获取方法及装置
CN113806464A (zh) * 2021-09-18 2021-12-17 北京京东乾石科技有限公司 路牙确定方法、装置、设备以及存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116503434A (zh) * 2023-05-15 2023-07-28 中国电建集团昆明勘测设计研究院有限公司 一种点云数据的边界提取方法、装置、设备及存储介质
CN116990832A (zh) * 2023-08-04 2023-11-03 长沙行深智能科技有限公司 危险道路边沿感知方法、装置、终端设备及存储介质

Also Published As

Publication number Publication date
CN113806464A (zh) 2021-12-17

Similar Documents

Publication Publication Date Title
WO2023040437A1 (zh) 路牙确定方法、装置、设备以及存储介质
CN111462275B (zh) 一种基于激光点云的地图生产方法和装置
Negru et al. Image based fog detection and visibility estimation for driving assistance systems
EP2806396A1 (en) Sparse light field representation
CN113362444A (zh) 点云数据的生成方法、装置、电子设备及存储介质
US20230005278A1 (en) Lane extraction method using projection transformation of three-dimensional point cloud map
EP3879484A2 (en) Satellite image processing method, network training method, related devices and electronic device
US11447153B2 (en) Method and apparatus for annotating virtual lane at crossing
US20210406564A1 (en) Perception data detection method and apparatus
WO2022237821A1 (zh) 生成交通标志线地图的方法、设备和存储介质
WO2023000221A1 (zh) 可行驶区域生成方法、可移动平台及存储介质
CN111696059B (zh) 一种车道线光滑连接处理方法和装置
CN114037966A (zh) 高精地图特征提取方法、装置、介质及电子设备
CN111507158A (zh) 利用语义分割检测停车区域的方法和装置
CN114443794A (zh) 数据处理和地图更新方法、装置、设备以及存储介质
CN112258519A (zh) 一种高精度地图制作中道路的让行线自动化提取方法及装置
CN112258568B (zh) 一种高精度地图要素的提取方法及装置
EP4080479A2 (en) Method for identifying traffic light, device, cloud control platform and vehicle-road coordination system
CN114005098A (zh) 高精地图车道线信息的检测方法、装置和电子设备
CN114972758A (zh) 一种基于点云弱监督的实例分割方法
WO2021189420A1 (zh) 一种数据处理方法及装置
WO2020135325A1 (zh) 移动设备定位方法、装置、系统及移动设备
CN101408977B (zh) 一种分割候选障碍物区域的方法及装置
US20230142243A1 (en) Device environment identification method and apparatus, electronic device, and autonomous vehicle
CN113920273B (zh) 图像处理方法、装置、电子设备和存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22868814

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 050/07/2024)