Map edge extraction method and device, electronic equipment and storage medium

Abstract

The invention relates to a map edge extraction method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the clustering of obstacles in a preset range based on a plurality of perception sensors, so as to generate a plurality of obstacle instances, discretizing the plurality of obstacle instances into a plurality of boundary coordinate sets in an initial grid map, discretely dividing the grid occupation probabilities of the boundary coordinate points in a first boundary coordinate set in the plurality of groups of boundary coordinate sets into a plurality of neighborhood grids of the boundary coordinate points, determining a second boundary coordinate set, and according to the first boundary coordinate set and the second boundary coordinate set, determining a plurality of neighborhood grids of the boundary coordinate points; and generating a first map edge line of the first boundary coordinate set corresponding to the first obstacle instance, and generating a map edge line in a preset range according to a plurality of edge lines of the plurality of obstacle instances. Therefore, the vehicle travelable area is determined in the form of the edge line, effective information is accurately output, and the vehicle can complete corresponding driving actions based on the edge line.

Description

Technical Field

[0001] The present disclosure relates to the field of autonomous driving technology, and in particular to a map edge extraction method, device, electronic device and storage medium. Background Art

[0002] In related technologies, the drivable area (Free Space) is the most important perception module in autonomous driving besides target detection. It is mainly applicable to the perception of obstacles (such as walls, parking lot pillars, curbs, etc.) that cannot be output in the form of target detection. At present, most of the solutions for Free Space detection fusion are to perform time-series accumulation of grid occupancy maps (Occupancy GridMap), maintain a grid image (Grid Map), and each grid (Cell) represents a certain size (Cell Size) of the object world. Each grid is filled with binary values ​​0 and 1. 0 means that the grid is not occupied, and 1 means that the grid is occupied. For example, a grid represents an area of ​​0.2m*0.2m in the physical world. The grid occupancy probability corresponding to the non-Free Space (driving area) area perceived by each sensor is accumulated in time series. Once a frame does not perceive an obstacle in this area, the probability is lowered. Finally, the grids with probability values ​​exceeding a certain threshold are regarded as grids occupied by obstacles and output. Summary of the invention

[0003] In order to overcome the problems existing in the related art, the present disclosure provides a map edge extraction method, device, electronic device and storage medium.

[0004] According to a first aspect of an embodiment of the present disclosure, a map edge extraction method is provided, comprising:

[0005] Performing obstacle clustering on obstacles within a preset range based on a plurality of perception sensors to generate a plurality of obstacle instances, wherein one obstacle instance includes row and column coordinates of a first grid occupied by an obstacle in an initial grid map, and a first grid occupancy probability of the first grid;

[0006] discretize the plurality of obstacle instances into a plurality of sets of boundary coordinate sets in the initial grid map according to the row and column coordinates of the plurality of obstacle instances;

[0007] Discretely divide the grid occupation probabilities of boundary coordinate points in a first boundary coordinate set in the multiple sets of boundary coordinate sets into multiple neighborhood grids of the boundary coordinate points, so as to determine a second boundary coordinate set according to a second grid occupied in the multiple neighborhood grids, wherein the first boundary coordinate set is any boundary coordinate set in the multiple sets of boundary coordinate sets;

[0008] Generate a first edge line of a first obstacle instance corresponding to the first boundary coordinate set according to the first boundary coordinate set and the second boundary coordinate set;

[0009] A map edge line within the preset range is generated according to the plurality of edge lines of the plurality of obstacle instances.

[0010] Optionally, the discretely dividing the grid occupancy probabilities of boundary coordinate points in a first boundary coordinate set in the multiple groups of boundary coordinate sets into a plurality of neighborhood grids of the boundary coordinate points, so as to determine a second boundary coordinate set according to an occupied second grid in the plurality of neighborhood grids, comprises:

[0011] Determine, according to the first boundary coordinate set and the first grid occupancy probability, a second grid occupancy probability of a boundary coordinate point in the first boundary coordinate set;

[0012] Performing Gaussian field sampling within a preset sampling radius of the boundary coordinate point of the first boundary coordinate set to discretely divide the second grid occupancy probability into a plurality of neighborhood grids of the boundary coordinate point, and obtaining a plurality of third grid occupancy probabilities of the plurality of neighborhood grids;

[0013] The neighboring grids whose occupation probabilities of the plurality of second grids are greater than a set threshold are taken as the occupied second grids;

[0014] The second boundary coordinate set is determined according to the row and column coordinates.

[0015] Optionally, performing Gaussian field sampling within a preset sampling radius of boundary coordinate points of the first boundary coordinate set includes:

[0016] Acquire a first perception accuracy of a first perception sensor, wherein the first boundary coordinate set is a boundary coordinate set of a first obstacle instance, and the first perception sensor is used to perceive the first obstacle instance;

[0017] Determining a first sampling radius of the boundary coordinate point according to the first perception accuracy, wherein the first perception accuracy is negatively correlated with the first sampling radius;

[0018] Gaussian field sampling is performed within the first sampling radius of the boundary coordinate point.

[0019] Optionally, clustering obstacles within a preset range based on multiple perception sensors to generate multiple obstacle instances includes:

[0020] Acquire the initial grid map, wherein the occupied third grid of the initial grid map is marked with a third grid occupancy probability;

[0021] performing obstacle clustering on the obstacles within the preset range based on the multiple perception sensors to generate multiple initial obstacle instances, wherein one of the initial obstacle instances includes a fourth grid occupancy probability of a fourth grid occupied by an obstacle in the initial grid map;

[0022] The plurality of obstacle instances are generated according to the third grid occupancy probability and the fourth grid occupancy probability.

[0023] Optionally, generating the plurality of obstacle instances according to the third grid occupancy probability and the fourth grid occupancy probability includes:

[0024] According to a preset attenuation rule, reducing the occupancy probability of the third grid;

[0025] superimposing the fourth grid occupancy probability onto the reduced third grid occupancy probability based on the row and column coordinates to generate the first grid occupancy probability;

[0026] The plurality of obstacle instances are generated according to the row and column coordinates and the first grid occupancy probability.

[0027] Optionally, generating the map edge line within the preset range according to the plurality of edge lines of the plurality of obstacle instances includes:

[0028] According to the positions of the multiple map edge lines in the initial grid map and the multiple obstacle instance categories corresponding to the multiple map edge lines, the multiple map edge lines are merged to generate map edge lines within the preset range.

[0029] Optionally, the fusing the plurality of edge lines according to the positions of the plurality of edge lines in the initial grid map and the plurality of obstacle instance categories corresponding to the plurality of map edge lines to generate the map edge lines within the preset range includes:

[0030] Determine a plurality of overlapping edge lines that meet a preset condition from the plurality of edge lines, wherein the preset condition is that the obstacle instances corresponding to the edge lines have the same category and overlap in position;

[0031] The map edge lines within the preset range are generated by fusing the multiple overlapping edge lines and the multiple edge lines.

[0032] Optionally, generating a first edge line of a first obstacle instance according to the first boundary coordinate set and the second boundary coordinate set includes:

[0033] Generate a hash table according to the first boundary coordinate set, where the hash table is used to map the row and column coordinates of the plurality of obstacle instances corresponding to the first grid;

[0034] Based on the second boundary coordinate set, updating the hash table;

[0035] The first edge line is generated according to the updated hash table.

[0036] According to a second aspect of an embodiment of the present disclosure, a map edge extraction device is provided, comprising:

[0037] A first generating module is configured to perform obstacle clustering on obstacles within a preset range based on a plurality of perception sensors to generate a plurality of obstacle instances, wherein one obstacle instance includes row and column coordinates of a first grid occupied by an obstacle in an initial grid map, and a first grid occupancy probability of the first grid;

[0038] A processing module, configured to discretize the plurality of obstacle instances into a plurality of sets of boundary coordinates in the initial grid map according to the row and column coordinates of the plurality of obstacle instances;

[0039] a determination module configured to discretely divide the grid occupancy probabilities of boundary coordinate points in a first boundary coordinate set in the plurality of boundary coordinate sets into a plurality of neighborhood grids of the boundary coordinate points, so as to determine a second boundary coordinate set according to a second grid occupied in the plurality of neighborhood grids, wherein the first boundary coordinate set is any boundary coordinate set in the plurality of boundary coordinate sets;

[0040] A second generating module is configured to generate a first edge line of the first boundary coordinate set corresponding to the first obstacle instance according to the first boundary coordinate set and the second boundary coordinate set;

[0041] The execution module is configured to generate a map edge line within the preset range according to the multiple edge lines of the multiple obstacle instances.

[0042] According to a third aspect of an embodiment of the present disclosure, a vehicle is provided, comprising:

[0043] processor;

[0044] a memory for storing processor-executable instructions;

[0045] The processor is configured to execute the executable instructions to implement the steps of any one of the methods described in the first aspect of the present disclosure.

[0046] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the program instructions are executed by a processor, the steps of the map edge extraction method provided in the first aspect of the present disclosure are implemented.

[0047] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0048] In the above manner, obstacles within a preset range are clustered based on a plurality of perception sensors to generate a plurality of obstacle instances, wherein one obstacle instance includes row and column coordinates of a first grid occupied by an obstacle in an initial grid map, and a first grid occupancy probability of the first grid; the plurality of obstacle instances are discretized into a plurality of groups of boundary coordinate sets in the initial grid map according to the row and column coordinates of the plurality of obstacle instances; the grid occupancy probabilities of boundary coordinate points in a first boundary coordinate set in the plurality of groups of boundary coordinate sets are discretized into a plurality of neighborhood grids of the boundary coordinate points; a second boundary coordinate set is determined according to a second grid occupied in the plurality of neighborhood grids; the first boundary coordinate set is any boundary coordinate set in the plurality of groups of boundary coordinate sets; a first map edge line corresponding to a first obstacle instance of the first boundary coordinate set is generated according to the first boundary coordinate set and the second boundary coordinate set; and a map edge line within a preset range is generated according to a plurality of edge lines of the plurality of obstacle instances. The impassable area is thus divided into different obstacle instances, and the position of the obstacle in the detection area is sent to the downstream in the form of instance edge lines, so that the downstream can determine the drivable area through the edge lines. The drivable area of ​​the vehicle is determined in the form of edge lines, and effective information is accurately output, so that the vehicle can complete the corresponding driving actions based on the edge lines.

[0049] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0051] Figure 1 The figure is a schematic diagram showing a grid occupancy map accumulation method according to an exemplary embodiment.

[0052] Figure 2 The figure is a flow chart of a map edge extraction method according to an exemplary embodiment.

[0053] Figure 3 It is a schematic diagram showing a method for determining a second boundary coordinate set according to an exemplary embodiment.

[0054] Figure 4 The figure is a block diagram of a map edge extraction device according to an exemplary embodiment.

[0055] Figure 5 is a block diagram of a vehicle 500 according to an exemplary embodiment. DETAILED DESCRIPTION

[0056] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0057] It should be noted that all actions of acquiring signals, information or data in this application are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0058] In the related art, the detection and fusion methods of the drivable area are mostly to accumulate the grid occupation map in time series. For example, Figure 1 is a schematic diagram of a grid occupancy map accumulation method according to an exemplary embodiment. Figure 1 As shown, the grid occupancy probability of each grid in the grid occupancy map of the vehicle's environment at time t0 is determined by detection, wherein the occupancy probability of obstacles in the grids of coordinates 1-3 is 0.6, the occupancy probability of obstacles in the grids of coordinates 1-4 is 0.8, and the occupancy probability of obstacles in the grids of coordinates 2-4 is 0.4; after time t0, obstacles in the environment are detected based on the sensors installed on the vehicle, and the occupancy probability of each grid in the vehicle's current environment perceived by the sensor is obtained, and the grid occupancy probability at time t0 and the grid occupancy probability detected by the sensor are superimposed to obtain the occupancy probability of obstacles in the current environment at time t1.

[0059] The grid occupancy probability determines that the grid with an obstacle probability greater than the set threshold is the grid occupied by the obstacle. Based on the judgment conditions, the position of the obstacle in the current environment can be determined. However, the obstacle position determined based on the above method has the following disadvantages:

[0060] (1) The boundary of the obstacle is not explicitly described, and the output is the overall grid position occupied by the obstacle. The output content is large, and the effective information used for the drivable area accounts for a low proportion;

[0061] (2) The perception results given by each sensor source are treated equally, without considering the differences in perception accuracy of different sensor sources, which leads to differences in obstacle detection accuracy.

[0062] In view of this, this proposal proposes a map edge extraction method. Figure 2 is a flow chart of a map edge extraction method according to an exemplary embodiment. Figure 2 As shown, the method is used in a vehicle-mounted terminal and includes the following steps.

[0063] In step S11, obstacles within a preset range are clustered based on multiple perception sensors to generate multiple obstacle instances, where one obstacle instance includes row and column coordinates of a first grid occupied by the obstacle in the initial grid map, and a first grid occupancy probability of the first grid.

[0064] The application scenario of this embodiment is described as an example. This embodiment is applied to a vehicle terminal, mainly in an autonomous driving scenario, but is not limited to autonomous driving. It can also be used to assist the driver in identifying environmental conditions, generate environmental information of the current vehicle environment, and assist the driver in completing driving tasks based on environmental information. During vehicle driving, it is necessary to identify obstacles in the surrounding environment to determine the drivable area of ​​the vehicle in the current environment, and complete the driving task based on the drivable area.

[0065] Obstacles within the preset range of the vehicle are sensed according to the sensing sensors arranged around the vehicle, wherein the sensing sensors may be lidar sensors, infrared sensors, camera sensors, etc., for sensing environmental information within the vehicle range in the corresponding direction. For example, in this embodiment, the sensor is arranged at a fixed position of the vehicle for detecting environmental information in a fixed direction. In order to obtain a full picture of environmental information within the preset range of the vehicle, multiple sensing sensors apply the principle of full coverage, and the corresponding detection angles can achieve 360° full coverage of the vehicle from a bird's eye view. The preset range can be adjusted according to the sensing detection requirements of the vehicle. For example, the preset range can be set to a square range of 6m with the center of the rear axle of the vehicle as the center point, and the environmental information within the square range is detected.

[0066] In this embodiment, it is necessary to determine the drivable area of ​​the vehicle in the current environment based on the environmental information. Therefore, multiple perception sensors arranged on the vehicle are used to detect obstacle information in the set direction. For example, obstacles within a preset range are clustered by multiple perception sensors to generate multiple obstacle instances. Obstacle clustering refers to the process of grouping and classifying obstacles in the environment according to their characteristics and positions. This clustering can help the vehicle better understand and process obstacle information in the environment. Obstacle instances corresponding to different obstacles are generated according to the clustering results of multiple perception sensors. In which, any obstacle instance includes the row and column coordinates of the first grid occupied by the obstacle in the initial grid map, and the first grid occupancy probability of the first grid. In this embodiment, the center of the rear axle of the vehicle is used as the center point to construct the initial grid map. The initial grid map (Occupancy Grid Map) is a commonly used map representation method in autonomous driving technology and autonomous navigation. It divides the environment into rectangular grids and models the occupancy state of each grid. The initial grid map is used to describe the distribution of obstacles in the area. Each grid can be marked as one of three states: occupied, free, or unknown. The occupied state indicates that the grid is occupied by an obstacle, the free state indicates that the grid is passable, and the unknown state indicates that there is a lack of information about the state of the grid. The grid occupancy map can be used to estimate the location information of objects in the environment and the passable area information. In the actual detection process, the occupancy state of the grid may be based on the position of the obstacle in the environment. There are differences in occupancy probabilities for the same obstacle. For example, the occupancy probability of the grid at the center of the obstacle is 1, and the occupancy probability of the grid at the edge of the obstacle can be any number between 0 and 1. The occupancy probability of the corresponding grid is determined based on the area occupied by the obstacle in the grid as sensed by the perception sensor. For example, the occupancy probability of the corresponding grid inside the obstacle is 1, and the occupancy probability of the corresponding grid at the edge of the obstacle is an approximate number between 0 and 1.

[0067] In the initial grid occupancy map, each grid is usually represented by a fixed size, which determines the resolution of the map to a certain extent. The higher the perception accuracy of the perception sensor, the higher the resolution of the map, and the smaller the grid, which can provide a more detailed description of the environment; the lower the perception accuracy of the perception sensor, the lower the resolution of the map, and the larger the grid. The grid occupancy map can be constructed and updated in a variety of ways, such as sensor data, laser scanning, or visual information. The initial grid map is generated by determining the occupancy probability of obstacles in each grid through the perception accuracy of multiple perception sensors in the initial state.

[0068] Optionally, in some embodiments, the above step S12 includes:

[0069] An initial grid map is obtained, wherein the occupied third grid in the initial grid map is marked with a third grid occupation probability;

[0070] Performing obstacle clustering on obstacles within a preset range based on a plurality of perception sensors to generate a plurality of initial obstacle instances, wherein an initial obstacle instance includes a fourth grid occupancy probability of a fourth grid occupied by an obstacle in the initial grid map;

[0071] A plurality of obstacle instances are generated according to the third grid occupancy probability and the fourth grid occupancy probability.

[0072] For example, an initial grid map is configured in the vehicle terminal, and the initial grid map is marked with historical obstacle instances within a preset range of the vehicle at historical moments. The historical obstacle instance is the probability of occupying the third grid in the grid map when the obstacle is sensed and determined by the vehicle through a sensor at the historical moment.

[0073] Obstacles within a preset range of the current vehicle are clustered according to multiple perception sensors to generate multiple initial obstacle instances, wherein one obstacle instance includes the fourth grid occupancy probability of the fourth grid occupied by the obstacle in the initial grid map, wherein the manner in which the sensors perform obstacle clustering is the same as that in the above step S11, and may refer to the above step S11 and will not be repeated herein.

[0074] The historical obstacle examples marked in the initial grid map and the initial obstacle instances sensed by the multiple perception sensors are superimposed with grid occupancy probabilities according to corresponding coordinate positions, thereby generating multiple obstacle instances, wherein one obstacle instance includes the row and column coordinates of the first grid occupied by the obstacle in the initial grid map, and the first grid occupancy probability of the first grid.

[0075] Optionally, in one implementation, the above step of “generating multiple obstacle instances according to the third grid occupancy probability and the fourth grid occupancy probability” includes:

[0076] According to the preset attenuation rule, the probability of occupying the third grid is reduced;

[0077] superimposing the fourth grid occupancy probability onto the reduced third grid occupancy probability based on the row and column coordinates to generate the first grid occupancy probability;

[0078] Generate multiple obstacle instances based on row and column coordinates and the first grid occupancy probability.

[0079] For example, in this embodiment, the third grid occupation probability is first reduced according to a preset attenuation rule. For example, an attenuation value can be set to attenuate the grid occupation probability of the occupied grid whose grid occupation probability in the initial grid map is not zero, and the attenuation value is subtracted from the third grid occupation probability to obtain the reduced third grid occupation probability. Among them, if the grid occupation probability after attenuation is a negative number, the grid occupation probability of the output third grid is 0. In this embodiment, the grid occupation probability in the initial grid map can be processed first, and the grids with grid occupation probability less than the threshold are marked as unoccupied grids, and the corresponding grid occupation probability is 0. Based on the processed initial grid map, the grid occupation probability is attenuated. For example, the occupied grids with a grid occupation probability lower than 0.125 are marked as non-occupied, and the corresponding grid occupation probability is 0.

[0080] The reduced third grid occupancy probability is superimposed with the fourth grid occupancy probability of the current vehicle environment sensed by multiple sensors according to the row and column coordinates of the grids in a bird's-eye view to generate a first grid occupancy probability of the occupied grid, and multiple obstacle instances are generated based on the first grid occupancy probability and the row and column coordinates of the first grid.

[0081] In step S12, the plurality of obstacle instances are discretized into a plurality of sets of boundary coordinates in the initial grid map according to the row and column coordinates of the plurality of obstacle instances.

[0082] For example, in this embodiment, based on the multiple obstacle instances generated in the above steps, the boundary grid position of each obstacle instance in the initial grid map is determined, and then based on the row and column coordinates of each obstacle instance occupying the initial grid map, the coordinate position of each obstacle instance boundary grid in the initial grid map is determined, and then the coordinate position is discretized into boundary coordinate points in the initial grid map, thereby generating multiple groups of boundary coordinate sets of each obstacle instance in the initial grid map. Among them, each obstacle instance corresponds to a group of boundary coordinate sets, and each group of boundary coordinate sets includes multiple boundary coordinate points. For example, the grid center coordinate point of the edge grid can be used as the coordinate position of the corresponding boundary coordinate point.

[0083] In step S13, the grid occupancy probability of the boundary coordinate points in the first boundary coordinate set in the multiple groups of boundary coordinate sets is discretely divided into multiple neighborhood grids of the boundary coordinate points, so as to determine the second boundary coordinate set according to the occupied second grid in the multiple neighborhood grids, and the first boundary coordinate set is any boundary coordinate set in the multiple groups of boundary coordinate sets.

[0084] In this embodiment, each boundary coordinate point corresponds to a boundary grid in the obstacle instance. In order to obtain a clearer boundary line corresponding to each obstacle instance, it is necessary to complete the boundary line of the obstacle instance based on the boundary coordinate set corresponding to the obstacle instance detected by the perception sensor, so as to generate a clear boundary line corresponding to each obstacle instance. For example, in this embodiment, the grid occupancy probability of the boundary grid in the initial grid map is discretely divided into multiple neighborhood grids adjacent to the boundary grid. According to the grid occupancy probability corresponding to each neighborhood grid, it is determined whether the neighborhood grid is a newly added boundary grid. In this embodiment, a probability threshold can be set for the grid occupancy probability, and the grid whose grid occupancy probability is greater than the set probability threshold is determined to be in an occupied state. According to this determination method, the newly occupied second grid is determined from multiple neighborhood grids, and then the coordinate position of the second boundary coordinate point corresponding to the second grid is determined based on the row and column coordinates to generate a second boundary coordinate set.

[0085] In step S14, a first map edge line corresponding to the first obstacle instance of the first boundary coordinate set is generated according to the first boundary coordinate set and the second boundary coordinate set.

[0086] For example, in this embodiment, the first boundary coordinate set and the second boundary coordinate set correspond to the same first obstacle instance, and the first boundary coordinate set and the second boundary coordinate set are merged to generate the first map edge line of the first obstacle instance. It can be understood that after the first boundary coordinate set and the second boundary coordinate set are merged, they are still in the form of points. The boundary coordinate points of the first obstacle instance in the initial grid map are described according to the coordinate point set. For example, the points in the point set can be connected in the way that adjacent points are connected, so as to generate the first map edge line corresponding to the first obstacle instance.

[0087] Optionally, in some implementations, the above step S14 includes:

[0088] Generate a hash table according to the first boundary coordinate set, where the hash table is used to map row and column coordinates of the first grid corresponding to the plurality of obstacle instances;

[0089] Based on the second boundary coordinate set, updating the hash table;

[0090] According to the updated hash table, a first edge line is generated.

[0091] For example, in this embodiment, a hash table is constructed based on the first boundary coordinate set, the row and column coordinates of the first grid occupied by multiple obstacle instances are mapped to the hash table, and the grid occupancy probability of the first grid is marked through the hash table. After determining the second boundary coordinate set through the above steps, the hash table is updated according to the second boundary coordinate set, and based on the updated hash table, the first edge line of the first obstacle instance is generated.

[0092] Through the above method, a hash table is used to record the row and column coordinates of the occupied grids and the grid occupancy probability in multiple obstacle instances. The hash table is updated based on the newly added grids, and the edge lines of the corresponding obstacle instances are output, thereby facilitating data recording and data updating and avoiding data confusion caused by data recording errors.

[0093] In step S15, a map edge line within a preset range is generated according to multiple edge lines of multiple obstacle instances.

[0094] For example, each obstacle instance is processed in the above manner to generate multiple edge lines corresponding to each obstacle instance, and the multiple edge lines are drawn in the same initial grid map to generate a map edge line within the current preset range of the vehicle.

[0095] In the above manner, obstacles within a preset range are clustered based on a plurality of perception sensors to generate a plurality of obstacle instances, wherein one obstacle instance includes row and column coordinates of a first grid occupied by an obstacle in an initial grid map, and a first grid occupancy probability of the first grid; the plurality of obstacle instances are discretized into a plurality of groups of boundary coordinate sets in the initial grid map according to the row and column coordinates of the plurality of obstacle instances; the grid occupancy probabilities of boundary coordinate points in a first boundary coordinate set in the plurality of groups of boundary coordinate sets are discretized into a plurality of neighborhood grids of the boundary coordinate points; a second boundary coordinate set is determined according to a second grid occupied in the plurality of neighborhood grids; the first boundary coordinate set is any boundary coordinate set in the plurality of groups of boundary coordinate sets; a first map edge line corresponding to a first obstacle instance of the first boundary coordinate set is generated according to the first boundary coordinate set and the second boundary coordinate set; and a map edge line within a preset range is generated according to a plurality of edge lines of the plurality of obstacle instances. The impassable area is thus divided into different obstacle instances, and the position of the obstacle in the detection area is sent to the downstream in the form of instance edge lines, so that the downstream can determine the drivable area through the edge lines. The drivable area of ​​the vehicle is determined in the form of edge lines, and effective information is accurately output, so that the vehicle can complete the corresponding driving actions based on the edge lines.

[0096] Optionally, in some implementations, the above step S15 includes:

[0097] According to the positions of the plurality of edge lines in the initial grid map and the plurality of obstacle instance categories corresponding to the plurality of edge lines, the plurality of map edge lines are fused to generate a map edge line within a preset range.

[0098] For example, in this embodiment, when multiple perception sensors perceive obstacles within a preset range, they can classify and identify the obstacles based on the perception algorithm to determine the category of each obstacle instance, and then obtain the edge line of each obstacle instance based on the above method, and merge the obstacle instances with the same obstacle instance category and overlapping edge lines to generate a map edge line within the preset range.

[0099] Optionally, in some implementations, the step of “merging the plurality of map edge lines according to their positions in the initial grid map and the plurality of obstacle instance categories corresponding to the plurality of map edge lines to generate a map edge line within a preset range” includes:

[0100] Determining a plurality of overlapping edge lines that meet a preset condition from a plurality of map edge lines, wherein the preset condition is that the obstacle instances corresponding to the edge lines have the same category and overlap in position;

[0101] A map edge line within a preset range is generated by fusing a plurality of overlapping map edge lines and a plurality of map edge lines.

[0102] For example, multiple obstacle instances are determined to meet the following conditions: the corresponding boundary lines overlap, and there are multiple overlapping edge lines with the same obstacle instance category, and then the overlapping edge lines are merged with other edge lines that do not overlap to form a map edge line within a preset range.

[0103] Figure 3 is a schematic diagram showing a method for determining a second boundary coordinate set according to an exemplary embodiment. Figure 3 As shown, in some implementations, the above step S13 may include the following steps.

[0104] In step S131 , the second grid occupancy probability of the boundary coordinate points in the first boundary coordinate set is determined according to the first boundary coordinate set and the first grid occupancy probability.

[0105] For example, in this embodiment, the second grid occupancy probability of the boundary coordinate point is the occupancy probability of the boundary grid corresponding to the boundary coordinate point in the initial grid map, the corresponding boundary grid is determined by the boundary coordinate position of the boundary coordinate point, the occupancy probability of the boundary grid is read, and the second grid occupancy probability of the boundary coordinate point is generated. The method for determining the second grid occupancy probability in this embodiment can be the same as in the above step S13, and can refer to the above step S13, and no further description is given.

[0106] In step S132, Gaussian field sampling is performed within a preset sampling radius of the boundary coordinate points of the first boundary coordinate set to discretely divide the second grid occupancy probability into multiple neighborhood grids of the boundary coordinate points to obtain multiple third grid occupancy probabilities of the multiple neighborhood grids.

[0107] For example, in this embodiment, the neighborhood grid around the boundary coordinate point is defined, and the complete grid within the preset sampling radius of the boundary coordinate point is determined as the neighborhood grid of the boundary coordinate point corresponding to the first boundary coordinate set. Different sampling radii are set for different accuracy requirements. For example, when it is determined that the vehicle currently requires a high accuracy for environmental detection, the preset sampling radius can be reduced to obtain fewer neighborhood grids, and multiple sampling methods are used to improve the sampling accuracy; when the vehicle currently requires a low accuracy for environmental detection, the preset sampling radius can be increased to obtain more neighborhood grids, and a few sampling extraction methods are used to quickly determine the occupied grids in the neighborhood grid. After the neighborhood grid of the boundary coordinate point is determined according to the preset sampling radius, the second grid occupancy probability of the boundary coordinate point is discretized by Gaussian field sampling and divided into multiple neighborhood grids of the boundary coordinate point, thereby obtaining multiple third grid occupancy probabilities of multiple neighborhood grids. Among them, in this embodiment, the grid occupancy probability of the boundary coordinate points of the first boundary coordinate set is based on Gaussian field sampling, and the Gaussian field model is used for modeling. By utilizing the statistical characteristics of the Gaussian field, the noise in the image can be effectively removed, and the third grid occupancy probability of the missing grid boundary coordinates can be estimated.

[0108] Optionally, in some implementations, the above step S132 includes:

[0109] Acquire a first perception accuracy of a first perception sensor, wherein the first boundary coordinate set is a boundary coordinate set of a first obstacle instance, and the first perception sensor is used to perceive the first obstacle instance;

[0110] Determine a first sampling radius of the boundary coordinate point according to the first perception accuracy, wherein the first perception accuracy is negatively correlated with the first sampling radius;

[0111] Gaussian field sampling is performed within the first sampling radius of the boundary coordinate points.

[0112] For example, in this embodiment, based on different types of perception sensors and perception requirements, the perception accuracy of each perception sensor configured on the vehicle is different. Therefore, it is necessary to fuse the boundary line of the obstacle instance based on the perception accuracy of each perception sensor. In this embodiment, the first perception sensor perceives environmental information to obtain the first obstacle instance, wherein the perception accuracy of each perception sensor is determined by predefining the perception parameters of the perception sensors configured in the vehicle, and the first perception accuracy of the first perception sensor in the vehicle terminal is determined according to the number of the first perception sensor.

[0113] The first sampling radius of the first boundary coordinate set of the first obstacle instance is determined according to the first perception accuracy. In this embodiment, the smaller the sampling radius is, the more sampling and extraction actions are performed based on the boundary coordinate points, and the corresponding newly added boundary coordinates are more accurate. Therefore, based on this characteristic, for the first perception sensor with a greater perception accuracy, the smaller the sampling radius of the boundary coordinate points of the corresponding first obstacle instance is, the more sampling times are, and the smaller the range of the determined neighborhood grid is; and the smaller the first perception accuracy is, the larger the corresponding sampling radius is, the more neighborhood grids are determined by the sampling radius, and the lower the accuracy of the newly added boundary grids is. For example, in this embodiment, the sampling radius and the perception accuracy are negatively correlated. The higher the perception accuracy of the perception sensor is, the smaller the corresponding sampling radius is. A negative correlation coefficient can be set, and the first perception accuracy is multiplied by the negative correlation coefficient to obtain the first sampling radius.

[0114] After determining the first sampling radius of the boundary coordinate points, Gaussian field sampling is performed on the grid occupancy probability of multiple boundary coordinate points within the first sampling radius. In this embodiment, the method of Gaussian field sampling after determining the sampling radius is the same as that in the above step S13, and the above steps can be referred to and will not be repeated.

[0115] Through the above method, based on the perception accuracy of different perception sensors, different sampling radii are set to sample the neighborhood grids around the boundary points, so as to fuse the perception of the sensors and make the newly added boundary coordinates more accurate.

[0116] In step S133, a plurality of neighborhood grids whose second grid occupation probabilities are greater than a set threshold are taken as occupied second grids.

[0117] For example, in this embodiment, the neighborhood grid whose second grid occupancy probability is greater than the set threshold is used as the newly occupied second grid. For example, the set threshold is determined to be 0.3, and the grid whose grid occupancy probability is greater than 0.3 is determined to be the occupied second grid. In this embodiment, the above-mentioned Gaussian field sampling process can be continuously repeated, and the sampling radius and the neighborhood grid are determined in the above-mentioned manner based on the newly added second grid, and the third grid occupancy probability of the second grid is discretely divided into multiple neighborhood grids, and the newly occupied fourth grid is determined according to the grid occupancy probability of the neighborhood grid. In this embodiment, iterative division can be performed based on the newly occupied grid until the number of iterations meets the preset number of iterations, and the newly occupied grid is used as the second grid.

[0118] In step S134, a second boundary coordinate set is determined according to the row and column coordinates.

[0119] For example, in this embodiment, the coordinate position of the newly added boundary coordinate point of the second grid is determined according to the row and column coordinates, and is merged with the coordinate position of the first boundary coordinate set to obtain the second boundary coordinate set.

[0120] Through the above method, the second grid occupancy probability is discretely divided into multiple neighborhood grids of boundary coordinate points by Gaussian neighborhood sampling with a preset sampling radius, and the third grid occupancy probability of the neighborhood grid is obtained, so as to determine the newly occupied boundary coordinate set based on the statistical characteristics of the Gaussian domain model.

[0121] Figure 4 is a block diagram of a map edge extraction device according to an exemplary embodiment. Figure 4 The device 100 includes: a first generating module 110, a processing module 120, a determining module 130, a second generating module 140 and an executing module 150.

[0122] The first generating module 110 is configured to perform obstacle clustering on obstacles within a preset range based on a plurality of perception sensors to generate a plurality of obstacle instances, wherein one obstacle instance includes row and column coordinates of a first grid occupied by an obstacle in an initial grid map, and a first grid occupancy probability of the first grid;

[0123] The processing module 120 is configured to discretize the plurality of obstacle instances into a plurality of sets of boundary coordinates in the initial grid map according to the row and column coordinates of the plurality of obstacle instances;

[0124] The determination module 130 is configured to discretely divide the grid occupation probabilities of the boundary coordinate points in a first boundary coordinate set in the plurality of boundary coordinate sets into a plurality of neighborhood grids of the boundary coordinate points, so as to determine a second boundary coordinate set according to a second grid occupied in the plurality of neighborhood grids, wherein the first boundary coordinate set is any boundary coordinate set in the plurality of boundary coordinate sets;

[0125] A second generating module 140 is configured to generate a first edge line of the first boundary coordinate set corresponding to the first obstacle instance according to the first boundary coordinate set and the second boundary coordinate set;

[0126] The execution module 150 is configured to generate a map edge line within the preset range according to the multiple edge lines of the multiple obstacle instances.

[0127] Optionally, the determination module 130 includes:

[0128] A first determining submodule is configured to determine a second grid occupancy probability of a boundary coordinate point in the first boundary coordinate set according to the first boundary coordinate set and the first grid occupancy probability;

[0129] A first processing submodule is configured to perform Gaussian field sampling within a preset sampling radius of a boundary coordinate point of the first boundary coordinate set, so as to discretely divide the second grid occupancy probability into a plurality of neighborhood grids of the boundary coordinate point, and obtain a plurality of third grid occupancy probabilities of the plurality of neighborhood grids;

[0130] A first generating submodule is configured to take the neighboring grids whose occupation probabilities of the plurality of second grids are greater than a set threshold as the occupied second grids;

[0131] The second determining submodule is configured to determine the second boundary coordinate set according to the row and column coordinates.

[0132] Optionally, the processing submodule is configured as follows:

[0133] Acquire a first perception accuracy of a first perception sensor, wherein the first boundary coordinate set is a boundary coordinate set of a first obstacle instance, and the first perception sensor is used to perceive the first obstacle instance;

[0134] Determining a first sampling radius of the boundary coordinate point according to the first perception accuracy, wherein the first perception accuracy is negatively correlated with the first sampling radius;

[0135] Gaussian field sampling is performed within the first sampling radius of the boundary coordinate point.

[0136] Optionally, the first generating module 110 includes:

[0137] An acquisition submodule is configured to acquire the initial grid map, wherein the occupied third grid of the initial grid map is marked with a third grid occupation probability;

[0138] a second generating submodule, configured to perform obstacle clustering on the obstacles within the preset range based on the multiple perception sensors to generate multiple initial obstacle instances, wherein one of the initial obstacle instances includes a fourth grid occupancy probability of a fourth grid occupied by an obstacle in the initial grid map;

[0139] The third generating submodule is configured to generate the plurality of obstacle instances according to the third grid occupancy probability and the fourth grid occupancy probability.

[0140] Optionally, the third generation submodule is configured as follows:

[0141] According to a preset attenuation rule, reducing the occupancy probability of the third grid;

[0142] superimposing the fourth grid occupancy probability onto the reduced third grid occupancy probability based on the row and column coordinates to generate the first grid occupancy probability;

[0143] The plurality of obstacle instances are generated according to the row and column coordinates and the first grid occupancy probability.

[0144] Optionally, the execution module 150 includes:

[0145] The second processing submodule is configured to fuse the multiple map edge lines according to their positions in the initial grid map and the multiple obstacle instance categories corresponding to the multiple map edge lines, so as to generate map edge lines within the preset range.

[0146] Optionally, the second processing submodule is configured to:

[0147] Determine a plurality of overlapping edge lines that meet a preset condition from the plurality of edge lines, wherein the preset condition is that the obstacle instances corresponding to the map edge lines have the same category and overlap in position;

[0148] The map edge lines within the preset range are generated by fusing the multiple overlapping edge lines and the multiple edge lines.

[0149] Optionally, the second generating module 140 is configured to:

[0150] Generate a hash table according to the first boundary coordinate set, where the hash table is used to map the row and column coordinates of the plurality of obstacle instances corresponding to the first grid;

[0151] Based on the second boundary coordinate set, updating the hash table;

[0152] The first edge line is generated according to the updated hash table.

[0153] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0154] The present disclosure also provides a computer-readable storage medium having computer program instructions stored thereon, and when the program instructions are executed by a processor, the steps of the map edge extraction method provided by the present disclosure are implemented.

[0155] Figure 5 is a block diagram of a vehicle 500 according to an exemplary embodiment. For example, the vehicle 500 may be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other types of vehicles. The vehicle 500 may be an autonomous vehicle, a semi-autonomous vehicle, or a non-autonomous vehicle.

[0156] Reference Figure 5, the vehicle 500 may include various subsystems, for example, an infotainment system 510, a perception system 520, a decision control system 530, a drive system 540, and a computing platform 550. The vehicle 500 may also include more or fewer subsystems, and each subsystem may include multiple components. In addition, each subsystem and each component of the vehicle 500 may be interconnected by wire or wireless means.

[0157] In some embodiments, the infotainment system 510 may include a communication system, an entertainment system, and a navigation system, among others.

[0158] The perception system 520 may include several sensors for sensing information about the environment around the vehicle 500. For example, the perception system 520 may include a global positioning system (the global positioning system may be a GPS system, or a Beidou system or other positioning systems), an inertial measurement unit (IMU), a laser radar, a millimeter wave radar, an ultrasonic radar, and a camera.

[0159] The decision control system 530 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0160] The drive system 540 may include components that provide powered motion for the vehicle 500. In one embodiment, the drive system 540 may include an engine, an energy source, a transmission system, and wheels. The engine may be an internal combustion engine, an electric motor, an air compression engine, or a combination of one or more thereof. The engine is capable of converting energy provided by the energy source into mechanical energy.

[0161] Some or all functions of the vehicle 500 are controlled by a computing platform 550. The computing platform 550 may include at least one processor 551 and a memory 552, and the processor 551 may execute instructions 553 stored in the memory 552.

[0162] The processor 551 may be any conventional processor, such as a commercially available CPU. The processor may also include a graphics processor (Graphic Process Unit, GPU), a field programmable gate array (Field Programmable Gate Array, FPGA), a system on chip (System on Chip, SOC), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC) or a combination thereof.

[0163] The memory 552 may be implemented by any type of volatile or nonvolatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0164] In addition to instructions 553 , memory 552 may also store data, such as road maps, route information, vehicle location, direction, speed, etc. The data stored in memory 552 may be used by computing platform 550 .

[0165] In the embodiment of the present disclosure, the processor 551 may execute the instruction 553 to complete all or part of the steps of the above-mentioned map edge extraction method.

[0166] In another exemplary embodiment, a computer program product is also provided. The computer program product includes a computer program executable by a programmable device. The computer program has a code portion for executing the above map edge extraction method when executed by the programmable device.

[0167] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the present disclosure. This application is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and examples are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0168] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A map edge extraction method, characterized in that: The method comprises: Performing obstacle clustering on obstacles within a preset range based on a plurality of perception sensors to generate a plurality of obstacle instances, wherein one obstacle instance includes row and column coordinates of a first grid occupied by an obstacle in an initial grid map, and a first grid occupancy probability of the first grid; discretize the plurality of obstacle instances into a plurality of sets of boundary coordinate sets in the initial grid map according to the row and column coordinates of the plurality of obstacle instances; Discretely divide the grid occupation probabilities of boundary coordinate points in a first boundary coordinate set in the multiple sets of boundary coordinate sets into multiple neighborhood grids of the boundary coordinate points, so as to determine a second boundary coordinate set according to a second grid occupied in the multiple neighborhood grids, wherein the first boundary coordinate set is any boundary coordinate set in the multiple sets of boundary coordinate sets; Generate a first edge line of a first obstacle instance corresponding to the first boundary coordinate set according to the first boundary coordinate set and the second boundary coordinate set; A map edge line within the preset range is generated according to the plurality of edge lines of the plurality of obstacle instances.

2. The method according to claim 1, characterized in that The step of discretely dividing the grid occupancy probabilities of boundary coordinate points in a first boundary coordinate set in the plurality of boundary coordinate sets into a plurality of neighborhood grids of the boundary coordinate points, and determining a second boundary coordinate set according to a second grid occupied in the plurality of neighborhood grids, comprises: Determine, according to the first boundary coordinate set and the first grid occupancy probability, a second grid occupancy probability of a boundary coordinate point in the first boundary coordinate set; Performing Gaussian field sampling within a preset sampling radius of the boundary coordinate point of the first boundary coordinate set to discretely divide the second grid occupancy probability into a plurality of neighborhood grids of the boundary coordinate point, and obtaining a plurality of third grid occupancy probabilities of the plurality of neighborhood grids; The neighboring grids whose occupation probabilities of the plurality of second grids are greater than a set threshold are taken as the occupied second grids; The second boundary coordinate set is determined according to the row and column coordinates.

3. The method according to claim 2, characterized in that The performing Gaussian field sampling within a preset sampling radius of the boundary coordinate points of the first boundary coordinate set includes: Acquire a first perception accuracy of a first perception sensor, wherein the first boundary coordinate set is a boundary coordinate set of a first obstacle instance, and the first perception sensor is used to perceive the first obstacle instance; Determining a first sampling radius of the boundary coordinate point according to the first perception accuracy, wherein the first perception accuracy is negatively correlated with the first sampling radius; Gaussian field sampling is performed within the first sampling radius of the boundary coordinate point.

4. The method according to claim 1, characterized in that: The obstacle clustering is performed on the obstacles within the preset range based on the multiple perception sensors to generate multiple obstacle instances, including: Acquire the initial grid map, wherein the occupied third grid of the initial grid map is marked with a third grid occupancy probability; performing obstacle clustering on the obstacles within the preset range based on the multiple perception sensors to generate multiple initial obstacle instances, wherein one of the initial obstacle instances includes a fourth grid occupancy probability of a fourth grid occupied by an obstacle in the initial grid map; The plurality of obstacle instances are generated according to the third grid occupancy probability and the fourth grid occupancy probability.

5. The method according to claim 4, characterized in that The generating the plurality of obstacle instances according to the third grid occupancy probability and the fourth grid occupancy probability comprises: According to a preset attenuation rule, reducing the occupancy probability of the third grid; superimposing the fourth grid occupancy probability onto the reduced third grid occupancy probability based on the row and column coordinates to generate the first grid occupancy probability; The plurality of obstacle instances are generated according to the row and column coordinates and the first grid occupancy probability.

6. The method according to claim 1, characterized in that The step of generating a map edge line within the preset range according to the plurality of edge lines of the plurality of obstacle instances comprises: According to the positions of the multiple map edge lines in the initial grid map and the multiple obstacle instance categories corresponding to the multiple map edge lines, the multiple map edge lines are merged to generate map edge lines within the preset range.

7. The method according to claim 6, characterized in that The step of fusing the plurality of edge lines according to the positions of the plurality of edge lines in the initial grid map and the plurality of obstacle instance categories corresponding to the plurality of map edge lines to generate the map edge lines within the preset range includes: Determining a plurality of overlapping edge lines that meet a preset condition from the plurality of edge lines, wherein the preset condition is that the obstacle instances corresponding to the edge lines have the same category and overlap in position; The map edge lines within the preset range are generated by fusing the multiple overlapping edge lines and the multiple edge lines.

8. The method according to any one of claims 1 to 7, characterized in that The step of generating a first edge line of the first obstacle instance according to the first boundary coordinate set and the second boundary coordinate set includes: Generate a hash table according to the first boundary coordinate set, where the hash table is used to map the row and column coordinates of the plurality of obstacle instances corresponding to the first grid; Based on the second boundary coordinate set, updating the hash table; The first edge line is generated according to the updated hash table.

9. A map edge extraction device, characterized in that: include: A first generating module is configured to perform obstacle clustering on obstacles within a preset range based on a plurality of perception sensors to generate a plurality of obstacle instances, wherein one obstacle instance includes row and column coordinates of a first grid occupied by an obstacle in an initial grid map, and a first grid occupancy probability of the first grid; A processing module, configured to discretize the plurality of obstacle instances into a plurality of sets of boundary coordinates in the initial grid map according to the row and column coordinates of the plurality of obstacle instances; a determination module configured to discretely divide the grid occupancy probabilities of boundary coordinate points in a first boundary coordinate set in the plurality of boundary coordinate sets into a plurality of neighborhood grids of the boundary coordinate points, so as to determine a second boundary coordinate set according to a second grid occupied in the plurality of neighborhood grids, wherein the first boundary coordinate set is any boundary coordinate set in the plurality of boundary coordinate sets; A second generating module is configured to generate a first edge line of the first boundary coordinate set corresponding to the first obstacle instance according to the first boundary coordinate set and the second boundary coordinate set; The execution module is configured to generate a map edge line within the preset range according to the multiple edge lines of the multiple obstacle instances.

10. A vehicle, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to execute the executable instructions to implement the steps of the method according to any one of claims 1 to 8.

11. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the program instructions are executed by a processor, the steps of the method described in any one of claims 1 to 8 are implemented.

Images