A method, device and equipment for determining a garage semantic map and a storage medium

Abstract

The application discloses a garage semantic map determination method, device and equipment and a storage medium. The method comprises the following steps: acquiring vehicle movement information of a target vehicle when the target vehicle moves along a planned path in a target garage and a parking space line image collected by a camera arranged on the target vehicle; performing semantic splicing on the parking space line image according to the vehicle movement information of the target vehicle to obtain an initial semantic map of the target garage; performing rasterization on the initial semantic map to obtain a pixel raster map, scoring pixel rasters in the pixel raster map, and determining raster scores of the pixel rasters; extracting an outer contour pixel raster set of each parking space line element in the pixel raster map according to the raster scores, and determining a target semantic map according to the outer contour pixel raster set. The application can reduce noise and redundant information existing in the semantic map, improve stability and accuracy of the semantic map, and facilitate storage and positioning loading.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of indoor positioning technology, and in particular to a garage semantic map determination method, device, equipment and storage medium. BACKGROUND

[0002] In order to realize long-distance automatic parking, it is necessary to establish a semantic map of the garage, and the parking line as the main marker on the ground of the garage can be used as the main reference for positioning in automatic parking of the vehicle.

[0003] Since the parking line semantics obtained by the perception pixel point has a lot of noise, the semantic map directly constructed by the parking line semantics obtained by the perception has a lot of noise and poor stability, which is not conducive to subsequent positioning; and the constructed semantic map has a lot of redundant information, which is not convenient for storage and positioning loading. SUMMARY

[0004] The present application provides a garage semantic map determination method, device, equipment and storage medium to solve the problem that the semantic map directly constructed by the parking line semantics obtained by the perception has a lot of noise and redundant information, reduce the noise and redundant information of the semantic map by the pixel grid scoring mechanism and the outer contour extraction method, improve the stability and accuracy of the semantic map, and facilitate storage and positioning loading.

[0005] According to an aspect of the present application, a garage semantic map determination method is provided, which comprises:

[0006] Obtaining vehicle motion information of a target vehicle when moving along a planned path in a target garage and a parking line image collected by a camera arranged on the target vehicle;

[0007] According to the vehicle motion information of the target vehicle, performing parking line semantic splicing on the parking line image to obtain an initial semantic map of the target garage;

[0008] Griding the initial semantic map to obtain a pixel grid map, scoring the pixel grid in the pixel grid map, and determining a grid score of the pixel grid;

[0009] According to the grid score, extracting an outer contour pixel grid set of each parking line element in the pixel grid map, and determining a target semantic map according to the outer contour pixel grid set.

[0010] According to another aspect of the present application, a garage semantic map determination device is provided, which comprises:

[0011] An information acquisition module is configured to acquire vehicle movement information of a target vehicle moving along a planned path in a target garage and parking line images collected by a camera arranged on the target vehicle;

[0012] A semantic splicing module is configured to perform semantic splicing on the parking line images according to the vehicle movement information of the target vehicle to obtain an initial semantic map of the target garage;

[0013] A grid scoring module is configured to rasterize the initial semantic map to obtain a pixel grid map, score pixel grids in the pixel grid map, and determine grid scores of the pixel grids;

[0014] A semantic map determination module is configured to extract a set of outer contour pixel grids of each parking line element in the pixel grid map according to the grid scores, and determine a target semantic map according to the set of outer contour pixel grids.

[0015] According to another aspect of the present application, an electronic device is provided, and the electronic device comprises:

[0016] at least one processor; and

[0017] a memory connected to the at least one processor in communication; wherein

[0018] the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the method for determining a garage semantic map according to any one of the embodiments of the present application.

[0019] According to another aspect of the present application, a computer readable storage medium is provided, and the computer readable storage medium stores computer instructions for enabling a processor to perform the method for determining a garage semantic map according to any one of the embodiments of the present application when the processor executes the computer instructions.

[0020] The technical scheme of the embodiment of the application obtains vehicle motion information of a target vehicle moving along a planned path in a target garage and a parking space line image collected by a camera arranged on the target vehicle; an initial semantic map of the target garage is obtained by performing semantic splicing of the parking space line image according to the vehicle motion information of the target vehicle; the initial semantic map is rasterized to obtain a pixel raster map, the pixel raster in the pixel raster map is scored to determine a raster score of the pixel raster; and an outer contour pixel raster set of each parking space line element in the pixel raster map is extracted according to the raster score, and a target semantic map is determined according to the outer contour pixel raster set. The pixel raster scoring mechanism is used to determine a raster with a lower raster score as a noise raster, which is not used as an extraction object of a parking space line element when the target semantic map is constructed, so as to reduce noise in the semantic map, improve stability and accuracy of the semantic map, extract an outer contour pixel of the pixel raster map corresponding to the initial semantic map by using an outer contour extraction method, reduce redundant information in the semantic map, so as to reduce an information amount of the semantic map, and facilitate storage and positioning loading of the semantic map.

[0021] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the application, nor is it used to limit the scope of the application. Other features of the application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0023] Figure 1 A flowchart of a garage semantic map determination method provided for the first embodiment of the application;

[0024] Figure 2 A schematic diagram of an initial semantic map provided for the first embodiment of the application;

[0025] Figure 3 A flowchart of a garage semantic map determination method provided for the second embodiment of the application;

[0026] Figure 4 A schematic diagram of a pixel raster map provided for the second embodiment of the application;

[0027] Figure 5 A schematic diagram of a camera coordinate system and a carrier coordinate system;

[0028] Figure 6A schematic diagram of a garage semantic map after noise elimination based on a score threshold value;

[0029] Figure 7 A flow chart of a garage semantic map determination method provided in Embodiment Three of the present application;

[0030] Figure 8 A schematic diagram of an outer contour of a parking line element;

[0031] Figure 9 A schematic diagram of a target semantic map determined according to an initial carrier coordinate of a pixel grid;

[0032] Figure 10 A schematic diagram of a method for determining a starting access pixel grid within a field of a current pixel grid;

[0033] Figure 11 A structural schematic diagram of a garage semantic map determination device provided in Embodiment Four of the present application;

[0034] Figure 12 A structural schematic diagram of an electronic device for implementing a garage semantic map determination method according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to make the personnel in the technical field better understand the present application scheme, the technical scheme in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the scope of protection of the present application.

[0036] It should be noted that the terms "first", "second", "target" and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include", "have", "contain" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0037] Embodiment One

[0038] Figure 1A flowchart of a method for determining a garage semantic map according to an embodiment of the present application is provided. The embodiment can be applied to a case of determining a garage semantic map. The method can be executed by a garage semantic map determination device, which can be implemented in the form of hardware and / or software. As shown in FIG. 46, the method comprises the following steps. Figure 1

[0039] In S110, vehicle motion information of a target vehicle moving along a planned path in a target garage and a parking space line image collected by a camera arranged on the target vehicle are acquired.

[0040] The target vehicle is a vehicle driving in the garage when the garage semantic map is determined. The garage semantic map can be understood as a map composed of semantic information of elements (such as parking space lines, lane lines, and indication information) contained in the garage, and can be used to guide automatic parking of a vehicle. The target garage is a garage for which the garage semantic map needs to be determined. The vehicle motion information is information generated by the vehicle during movement, which can include inertial navigation information and speed information, for example. The parking space line image is an image collected by the camera, which contains part or all of the semantic information of the parking space lines.

[0041] Specifically, the vehicle motion information can be acquired by controlling the target vehicle to move along the planned path in the target garage, and acquiring the vehicle motion information of the vehicle through a vehicle host system of the vehicle during movement of the target vehicle. Alternatively, the vehicle motion information can be acquired by controlling the target vehicle to move along the planned path in the target garage, and acquiring inertial navigation information of the target vehicle through an inertial navigation device (such as a gyroscope and an accelerometer) on the target vehicle, and acquiring speed information of the target vehicle through a wheel speed sensor on the target vehicle.

[0042] The parking space line image can be acquired by arranging a camera on the target vehicle, and collecting the parking space line image at a preset shooting frequency during movement of the target vehicle along the planned path. Since the field of view of the camera is limited, the target vehicle is controlled to move around the garage for one round along the planned path, so that the camera on the target vehicle can collect images of all parking space lines in the target garage. The mounting position of the camera on the target vehicle is not limited in the embodiment of the present application, and can be set according to actual conditions. Generally, the camera can be arranged on the top of the vehicle and face the front of the vehicle.

[0043] In S120, an initial semantic map of the target garage is obtained by performing parking space line semantic splicing on the parking space line image according to the vehicle motion information of the target vehicle.

[0044] The initial semantic map is a semantic map obtained by performing parking space line splicing on the parking space line image collected by the camera. The initial semantic map can contain a large amount of noise and redundant information.

[0045] ​Specifically, the collection positions corresponding to each parking line image are determined according to the target vehicle motion information, and the initial semantic map of the target garage can be obtained by stitching the parking lines according to the coordinates of the pixels in the parking line image in the camera coordinate system corresponding to each collection position. Figure 2

[0046] Specifically, since the camera coordinate system of each collection position can be different, the camera coordinate systems corresponding to different collection positions can be converted into the same coordinate system, and then the parking lines can be stitched according to the coordinates of the pixels in the parking line image. The method for converting different camera coordinate systems into a unified coordinate system is not limited in the embodiments of the present application.

[0047] S130, rasterizing the initial semantic map to obtain a pixel raster map, scoring the pixel raster in the pixel raster map, and determining the raster score of the pixel raster.

[0048] The pixel raster map is the rasterized map of the initial semantic map. The raster score is the score of the pixel raster after scoring.

[0049] Specifically, the initial semantic map is rasterized according to a certain size to obtain a pixel raster map, and the pixel raster is scored according to a preset scoring mechanism to obtain the raster score of each pixel raster. For example, the preset scoring mechanism can score according to the proportion of the parking line pixel points in the pixel raster to all the pixel points in the raster. The higher the proportion of the parking line pixel points in the raster, the higher the raster score. The preset scoring mechanism can also score according to the number of parking line pixel points contained in the pixel raster. The more the number of parking line pixel points contained in the pixel raster, the higher the raster score.

[0050] For example, the size of the raster can be selected according to actual needs, for example, it can be 10cm*10cm.

[0051] S140, extracting the outer contour pixel raster set of each parking line element in the pixel raster map according to the raster score, and determining the target semantic map according to the outer contour pixel raster set.

[0052] The parking line element is the main factor constituting the parking line. Since the camera cannot guarantee to completely collect all the images of the parking line in the process of collecting the parking line image, the pixel set containing the parking line element can be regarded as a parking line. The outer contour pixel raster set of the parking line element is a set composed of the pixel rasters corresponding to the outer contour of the parking line element. The target semantic map is the final determined garage semantic map; the target semantic map determined by the embodiments of the present application is a semantic map after reducing noise and redundant information.

[0053] ​Specifically, the way of extracting the outer contour pixel grid set of each parking line element in the pixel grid map according to the grid score can be inputting the pixel grid map and the grid score of each corresponding grid into a trained complete contour extraction model to obtain the outer contour pixel grid set corresponding to the parking line element, and the trained complete contour extraction model can be trained according to a pixel grid map training sample containing a semantic label of the grid score; or the grid score greater than a preset threshold in the pixel grid map can be determined, and the outer contour pixel grid set of each parking line element in the grid with the grid score greater than the preset threshold can be extracted.

[0054] In this step, the grid with a lower grid score is determined as a noise grid through the pixel grid scoring mechanism, and is not used as an extraction object of the parking line element when the target semantic map is constructed, so as to reduce the noise existing in the semantic map and improve the stability and accuracy of the semantic map; through the outer contour extraction method, the outer contour pixel of the pixel grid map corresponding to the initial semantic map is extracted, the redundant information existing in the semantic map is reduced, so as to reduce the information amount of the semantic map, and facilitate the storage and positioning loading of the semantic map.

[0055] In this step, the grid with a lower grid score is determined as a noise grid through the pixel grid scoring mechanism, and is not used as an extraction object of the parking line element when the target semantic map is constructed, so as to reduce the noise existing in the semantic map and improve the stability and accuracy of the semantic map; through the outer contour extraction method, the outer contour pixel of the pixel grid map corresponding to the initial semantic map is extracted, the redundant information existing in the semantic map is reduced, so as to reduce the information amount of the semantic map, and facilitate the storage and positioning loading of the semantic map.

[0056] Embodiment two

[0057] Figure 3A flowchart of a garage semantic map determination method provided for the second embodiment of the present application is based on the step S130 of the above-mentioned embodiment. The step S130, rasterizing the initial semantic map to obtain a pixel raster map, scoring the pixel raster in the pixel raster map, and determining the raster score of the pixel raster, comprises: rasterizing the initial semantic map based on a preset size of the raster to obtain a pixel raster map of the initial semantic map; determining the pixel raster coordinates corresponding to each parking line pixel point in the pixel raster map according to the initial carrier coordinates of the parking line pixel points in the initial semantic map; determining the number of parking line pixel points contained by each pixel raster according to the pixel raster coordinates corresponding to each parking line pixel point; and determining the raster score of the pixel raster according to the number of pixel points in the pixel raster. As shown in Figure 3 the method comprises:

[0058] S210, obtaining vehicle motion information of a target vehicle when the target vehicle moves along a planned path in a target garage and a parking line image collected by a camera arranged on the target vehicle.

[0059] S220, performing parking line semantic splicing on the parking line image according to the vehicle motion information of the target vehicle to obtain an initial semantic map of the target garage.

[0060] S230, rasterizing the initial semantic map based on a preset size of the raster to obtain a pixel raster map of the initial semantic map.

[0061] The preset size of the raster can be set according to the actual situation of the target garage. It can be understood that the smaller the raster, the more accurate the parking line semantic information of the determined target semantic map, but the larger the data amount of the target semantic map, the larger the calculation amount of the processing process of the initial semantic map, and the higher the requirement for hardware resources. If the raster is too large, it may cause too much parking line semantic information to be filtered, resulting in an inaccurate target semantic map. Exemplarily, the preset size of the raster can be set to 10cm*10cm.

[0062] Specifically, the initial semantic map is rasterized according to the preset size of the raster to obtain the pixel raster map of the initial semantic map. Exemplarily, the initial semantic map shown in Figure 2 is rasterized to obtain the pixel raster map of the initial semantic map as shown in Figure 4 .

[0063] S240, determining the pixel raster coordinates corresponding to each parking line pixel point in the pixel raster map according to the initial carrier coordinates of the parking line pixel points in the initial semantic map.

[0064] Wherein, the parking line pixel point is the smallest image unit representing the parking line position in the parking line image. The initial carrier coordinate is the coordinate of the parking line pixel point in the carrier coordinate system at the initial time. The carrier coordinate system can be a coordinate system established with the projection point of the center of the front bumper of the vehicle to the ground as the origin, with the right side of the driving direction of the vehicle as the positive direction of the X axis, with the driving direction of the vehicle as the positive direction of the Y axis, and with the direction perpendicular to the XOY plane downward as the positive direction of the Z axis. The origin of the carrier coordinate system is located on the ground, and the output parking line position is a two-dimensional coordinate on the ground, so the z component is 0. The pixel grid coordinate is the coordinate of the pixel grid in the initial carrier coordinate system.

[0065] Specifically, the maximum and minimum values of the x and y coordinates of the parking line pixel point are determined according to the initial carrier coordinate of the parking line pixel point, the width and height of the pixel grid map are calculated according to the maximum and minimum values of the x and y coordinates of the parking line pixel point, and the pixel grid coordinate corresponding to each parking line pixel point is determined according to the initial carrier coordinate of each parking line pixel point, the minimum value of the x and y coordinates of the parking line pixel point, and the width and height of the pixel grid.

[0066] For example, the pixel grid coordinate corresponding to the parking line pixel point is determined by the following steps:

[0067] (1) The maximum and minimum values of the x and y coordinates of the parking line pixel point are determined according to the initial carrier coordinate of the parking line pixel point. min , x max , y min , y max , the width w and the height h of the pixel grid map are calculated:

[0068] w = floor((x max -x min ) / w c +1);

[0069] h = floor((y max -y min ) / h c +1)

[0070] Wherein, floor is used to calculate the maximum integer not greater than a floating point number, w c and h c represent the width and height of the grid, respectively.

[0071] (2) The pixel grid coordinates u, v of each parking line pixel point [x y 0] T are calculated respectively:

[0072] u = floor((x-x min ) / w c );

[0073] v = floor((y-ymin ) / h c )

[0074] The width and height of the pixel grid are set according to actual needs. For example, the width and height of the pixel grid can be w c =h c =0.01m.

[0075] S250, according to the pixel grid coordinates corresponding to each parking line pixel point, determine the number of parking line pixel points contained in each pixel grid.

[0076] Specifically, according to the pixel grid coordinates corresponding to each parking line pixel point, determine the pixel grid to which each parking line pixel point belongs, and according to the pixel grid to which each parking line pixel point belongs, determine the number of parking line pixel points contained in each pixel grid.

[0077] S260, determine the grid score of the pixel grid according to the number of pixel points in the pixel grid.

[0078] Specifically, determining the grid score of the pixel grid according to the number of pixel points in the pixel grid can be directly taking the number of pixel points as the grid score of the pixel grid, or determining the grid score of the pixel grid according to the corresponding relationship between the number of pixel points in the pixel grid and the grid score.

[0079] For example, the corresponding relationship between the number of pixel points in the pixel grid and the grid score is established in advance, and after determining the number of pixel points in the pixel grid, the grid score corresponding to the number of pixel points in the pixel grid is obtained by looking up the corresponding relationship established in advance. The corresponding relationship between the number of pixel points and the grid score can be that each level of grid score corresponds to a pixel point quantity interval, for example, the pixel point quantity interval {0, 1000} corresponds to the grid score 0; the pixel point quantity interval {1001, 2000} corresponds to the grid score 1, the pixel point quantity interval {2001, 3000} corresponds to the grid score 2, and so on.

[0080] S270, extract the outer contour pixel grid set of each parking line element in the pixel grid map according to the grid score, and determine the target semantic map according to the outer contour pixel grid set.

[0081] The embodiment of the application performs rasterization on an initial semantic map based on a preset size grid to obtain a pixel grid map of the initial semantic map; determines pixel grid coordinates corresponding to each parking line pixel point in the pixel grid map according to initial carrier coordinates of the parking line pixel points in the initial semantic map; determines the number of parking line pixel points contained in each pixel grid according to the pixel grid coordinates corresponding to each parking line pixel point; and determines a grid score of the pixel grid according to the number of pixel points in the pixel grid. The grid score of the pixel grid is determined according to the number of pixel points in the pixel grid, and the pixel grid can be scored simply and quickly.

[0082] Optionally, the initial semantic map of the target garage is obtained by performing parking line semantic splicing on the parking line image according to the vehicle motion information of the target vehicle, and includes the following steps.

[0083] A carrier coordinate system corresponding to each image acquisition time of the target vehicle in the motion process is created;

[0084] The parking line pixel points on the parking line image acquired at each image acquisition time are determined, and the acquisition time carrier coordinates of the parking line pixel points in the corresponding acquisition time carrier coordinate system are determined;

[0085] For each parking line image, the initial carrier coordinates of the acquisition time carrier coordinates of the parking line pixel points in the initial carrier coordinate system are determined according to the vehicle motion information of the target vehicle, and the initial carrier coordinate system is the carrier coordinate system corresponding to the initial motion position of the target vehicle;

[0086] The parking line pixel points in the parking line image in the initial carrier coordinate system are subjected to parking line semantic splicing to obtain the initial semantic map of the target garage.

[0087] The acquisition time carrier coordinate system is a carrier coordinate system established according to the position of the target vehicle at the acquisition time. The acquisition time carrier coordinates are the coordinates of the parking line pixel points in the corresponding acquisition time carrier coordinate system. The initial carrier coordinate system is the carrier coordinate system of the target vehicle at the starting position. For example, the acquisition time carrier coordinate system can be a plane rectangular coordinate system with the center of the front bumper of the target vehicle at the acquisition time as the origin and the vehicle motion direction as the positive direction of the Y axis.

[0088] Specifically, at each image acquisition time in the motion process of the target vehicle, the parking line image is acquired and the acquisition time carrier coordinate system is created, and the pixel coordinates of the parking line pixel points in the parking line image coordinate system in the parking line image can be obtained. According to the camera intrinsic parameters and the camera extrinsic parameters, the pixel coordinates of the parking line pixel points in the parking line image coordinate system are converted to the acquisition time carrier coordinates in the acquisition time carrier coordinate system through coordinate conversion.

[0089] However, since the carrier coordinate system at the collection time is established according to the position of the vehicle at the collection time, the positions of the target vehicle at different collection times can be different, and the established carrier coordinate system can also be different. Therefore, the carrier coordinates at the collection time corresponding to the parking space line pixel points on the parking space line image collected at different collection times are obtained in the initial carrier coordinate system through the second coordinate conversion according to the vehicle motion information of the target vehicle; and the initial semantic map of the target garage is obtained by performing parking space line semantic splicing according to the initial carrier coordinates of the parking space line pixel points in the initial carrier coordinate system in each parking space line image.

[0090] For example, according to the vehicle motion information of the target vehicle, the manner of determining the initial carrier coordinates of the carrier coordinates at the collection time of the parking space line pixel points in the initial carrier coordinate system can be as follows:

[0091] According to the vehicle motion information of the target vehicle, the pose matrix T of the carrier coordinate system at the collection time relative to the carrier coordinate system at the initial time is determined b , and the carrier coordinates at the collection time of the parking space line pixel points are p b ; according to the pose matrix T of the carrier coordinate system at the collection time relative to the carrier coordinate system at the initial time b and the carrier coordinates at the collection time of the parking space line pixel points being p b , the initial carrier coordinates p of the carrier coordinates at the collection time of the parking space line pixel points in the initial carrier coordinate system are determined w , that is:

[0092]

[0093]

[0094] , wherein T b is the pose matrix of the carrier coordinate system at the collection time relative to the carrier coordinate system at the initial time, R b is the attitude matrix of the carrier system relative to the initial time, t b is the displacement amount of the carrier system relative to the initial time, p b is the carrier coordinates at the collection time of the parking space line pixel points, and p w is the initial carrier coordinates of the carrier coordinates at the collection time of the parking space line pixel points in the initial carrier coordinate system.

[0095] Optionally, the carrier coordinates at the collection time of the parking space line pixel points on the parking space line image collected at each image collection time in the corresponding carrier coordinate system at the collection time include:

[0096] The distortion correction is performed on the parking space line image collected at each image collection time.

[0097] The pixel points of the parking space lines are perceived to obtain the pixel points corresponding to each parking space line;

[0098] According to the intrinsic and extrinsic parameters of the camera, the collection time carrier coordinates of the pixel points of the parking space lines in the corresponding collection time carrier coordinate system are determined.

[0099] The extrinsic parameters of the camera include a transformation matrix of the camera coordinate system of the camera and the corresponding collection time carrier coordinate system.

[0100] The distortion correction is to correct the deviation of the pixel points in the collected parking space line image caused by the lens precision. The intrinsic parameters of the camera are a parameter matrix for conversion from the camera coordinate system image to the image coordinate system. The extrinsic parameters of the camera are a parameter matrix for conversion from the camera coordinate system to the carrier coordinate system.

[0101] Specifically, the parking space line image collected at each image collection time is corrected by a distortion correction model. The pixel points of the parking space lines corresponding to each parking space line are obtained by perceiving the pixel points of the parking space lines after distortion correction. According to the intrinsic parameters of the camera, the coordinates of the pixel points of the parking space lines in the image coordinate system are converted to the coordinates of the camera coordinate system. According to the extrinsic parameters of the camera, the coordinates of the pixel points of the parking space lines in the camera coordinate system are converted to the coordinates of the carrier coordinate system.

[0102] Exemplarily, Figure 5 is a schematic view of the camera coordinate system and the carrier coordinate system. As shown in Figure 5 , B is the carrier coordinate system, and C is the camera coordinate system. According to the intrinsic and extrinsic parameters of the camera, the pixel coordinates of the pixel points of the parking space lines in the parking space line image coordinate system are converted to the collection time carrier coordinates in the collection time carrier coordinate system in the following manner:

[0103] Let the coordinates of the pixel points of the parking space lines in the image coordinate system be where u represents the horizontal coordinate of the pixel points of the parking space lines on the parking space line image, v represents the vertical coordinate of the pixel points of the parking space lines on the parking space line image, and the top left corner of the parking space line image is taken as the origin of the image coordinate system, and the odd power form is:

[0104]

[0105] The coordinates p i of the pixel points of the parking space lines in the image coordinate system are converted to the coordinates p c of the camera coordinate system in the following manner according to the intrinsic and extrinsic parameters of the camera:

[0106] sp i = Kp c ;

[0107] n = [n1 n2 n3 n4] = [0 0 1 0]T bc ;

[0108] where s represents a scale factor, K is a camera intrinsic matrix, p c is a coordinate of a parking line pixel point in a camera coordinate system, p c is a 3-dimensional column vector; T bc is a camera extrinsic, and n is a ground parameter, n is a 4-dimensional row vector, which consists of four components n1, n2, n3, and n4.

[0109] The ground equation constraint of the coordinate p c of the parking line pixel point in the camera coordinate system is:

[0110] [n1 n2 n3]p c +n4 = 0.

[0111] By simultaneously solving and eliminating the scale factor s, the coordinate p c of the parking line pixel point in the camera coordinate system can be obtained:

[0112]

[0113] According to the camera extrinsic T bc , the coordinate p c of the parking line pixel point in the camera coordinate system can be converted into the coordinate p b in the carrier coordinate system in the following manner:

[0114]

[0115] p b = R bc p c + t bc ;

[0116] where p b is a carrier coordinate of the parking line pixel point at a collection time, T bc is a camera extrinsic, R bc is a 3x3 rotation matrix, and t bc is a 3x1 translation vector.

[0117] Optionally, if the grid score of the pixel grid is less than a score threshold, the grid score of the pixel grid is set to zero.

[0118] where the score threshold is determined according to the number of pixel grids with a non-zero grid score in the pixel grid map, the preset size of the grid, and the perception error rate of the pixel point.

[0119] where the perception error rate of the pixel point is the probability of error when identifying the parking line pixel point from the parking line image.

[0120] Specifically, for the pixel grids with non-zero grid scores, the pixel grids are sorted according to the grid scores from small to large; a plurality of grid scores are obtained according to the product of the number of pixel grids with non-zero grid scores in the pixel grid map, the preset size of the grid and the perception error rate of the pixel point; the grid scores of the pixel grids with lower grid scores in the plurality of pixel grids at the front of the sorting are set to zero; and the grid score of the last pixel grid set to zero is taken as the score threshold.

[0121] For example, it is assumed that the number of pixel grids with non-zero grid scores in the pixel grid map is N, the preset size of the grid is 0.01 m 2 , and the perception error rate of the pixel point is δ, and the score threshold S is S = 0.01 × N × δ. For example, the grid scores of the pixel grids less than the score threshold in the pixel grid map shown in FIG. 6 are set to zero, the noise grids in the pixel grid map are eliminated, and the garage semantic map after eliminating the noise is obtained as shown in FIG. 7. Figure 4 Figure 6

[0122] In this step, by setting the grid scores of the pixel grids with grid scores less than the score threshold to zero, the noise can be effectively filtered, and the stable parking line element is retained.

[0123] Embodiment Three

[0124] Figure 7 A flowchart of a method for determining a garage semantic map is provided in the third embodiment of the present application, and the present embodiment is refined on the basis of step S140 of the above-mentioned embodiments. Step S140, extracting the outer contour pixel grid set of each parking line element in the pixel grid map according to the grid score, determining the target semantic map according to the outer contour pixel grid set, comprises: determining the parking line element in the pixel grid map according to the grid score of each pixel grid in the pixel grid map; the parking line element is a connected domain composed of a plurality of pixel grids; extracting the outer contour pixel grid set of each parking line element according to the grid coordinates of each pixel grid in the pixel grid map; determining the initial carrier coordinates of the pixel grid coordinates corresponding to the pixel grid in the outer contour pixel grid set in the initial carrier coordinate system; determining the target semantic map according to the initial carrier coordinates of the pixel grids in each outer contour pixel grid set. As shown in FIG. 8, the method comprises: Figure 7

[0125] S310, obtaining the vehicle motion information of the target vehicle when moving along the planned path in the target garage and the parking line image collected by the camera arranged on the target vehicle.

[0126] S320, according to the vehicle motion information of the target vehicle, performing parking line semantic splicing on the parking line image to obtain an initial semantic map of the target garage. ​​​

[0127] S330, rasterizing the initial semantic map to obtain a pixel grid map, scoring the pixel grids in the pixel grid map, and determining a grid score of the pixel grids.

[0128] S340, determining a parking line element in the pixel grid map according to the grid score of each pixel grid in the pixel grid map; the parking line element is a connected domain composed of a plurality of pixel grids.

[0129] Specifically, the manner of determining the parking line element in the pixel grid map according to the grid score of each pixel grid in the pixel grid map can be that, for the pixel grid map after eliminating noise by the zeroing manner, the connected domain composed of the pixel grids with a grid score greater than zero in the pixel grid map is determined as a parking line element; or for the pixel grid map without eliminating noise by the zeroing manner, the connected domain composed of the pixel grids with a grid score greater than a score threshold in the pixel grid map is determined as a parking line element, and in the process of determining the parking line element, the pixel grids with a grid score lower than the score threshold are excluded and are not considered, so that the purpose of eliminating noise can also be achieved.

[0130] S350, extracting an outer contour pixel grid set of each parking line element according to the grid coordinates of each pixel grid in the pixel grid map.

[0131] Specifically, the position of each pixel grid is determined according to the grid coordinates of each pixel grid in the pixel grid map, and the outer contour pixel grid set of the parking line element is extracted according to the position of the pixel grid to obtain a schematic diagram of the outer contour of the parking line element as shown in Figure 8 .

[0132] S360, determining an initial carrier coordinate of the pixel grid in the outer contour pixel grid set in the initial carrier coordinate system.

[0133] Specifically, the initial carrier coordinate of the pixel grid is determined according to the pixel grid coordinates corresponding to the pixel grids in the outer contour pixel grid set, the width and height of the pixel grid, and the maximum and minimum values of the x and y coordinates of the pixel points of the parking line.

[0134] S370, determining a target semantic map according to the initial carrier coordinates of the pixel grids in each outer contour pixel grid set.

[0135] Specifically, the target semantic map as shown in Figure 9 is obtained by splicing the pixel grids according to the initial carrier coordinates of the pixel grids in each outer contour pixel grid set.

[0136] In this embodiment, the parking line element in the pixel grid map is determined according to the grid score of each pixel grid in the pixel grid map; the parking line element is a connected domain composed of a plurality of pixel grids; the outer contour pixel grid set of each parking line element is extracted according to the grid coordinates of each pixel grid in the pixel grid map; the initial carrier coordinates of the pixel grid in the outer contour pixel grid set in the initial carrier coordinate system are determined; and the target semantic map is determined according to the initial carrier coordinates of the pixel grid in each outer contour pixel grid set. The method can simplify the extraction of the outer contour pixel grid set of each parking line element and improve the contour extraction efficiency.

[0137] Optionally, in step S340, the parking line element in the pixel grid map is determined according to the grid score of each pixel grid in the pixel grid map, and the step S340 comprises:

[0138] S341, selecting a pixel grid to be accessed from the pixel grid map; the pixel grid to be accessed is a pixel grid with a grid score greater than a preset threshold and not yet accessed.

[0139] In the formula, the pixel grid to be accessed is the pixel grid currently waiting to be accessed. The process queue is a queue for storing the current access pixel grid being accessed; and the output queue is a queue for storing the pixel grid marked as having been accessed.

[0140] Specifically, a pixel grid with a grid score greater than a preset threshold and not yet accessed is randomly selected from the pixel grid map as the pixel grid to be accessed. If the pixel grid with a grid score less than the score threshold is zeroed in advance, the preset threshold can be zero; if the pixel grid with a grid score less than the score threshold is not zeroed, the preset threshold can be the score threshold.

[0141] S342, creating an output queue and a process queue.

[0142] Specifically, the output queue is created to store the pixel grid to be accessed, and the process queue is created to store the pixel grid marked as having been accessed.

[0143] S343, writing the pixel grid to be accessed into the process queue; and writing the current access pixel grid determined from the pixel grid to be accessed in the process queue into the output queue.

[0144] Specifically, the selected pixel grid to be accessed is written into the process queue, the current access pixel grid is determined in turn according to the writing order of the pixel grid to be accessed in the process queue, and each current access pixel grid determined at a time is written into the output queue to access the current access pixel grid in the output queue.

[0145] It can be understood that the number of the pixel grids to be accessed selected from the pixel grid map in each cycle is one, and thus the pixel grid to be accessed is written into the output queue. If the pixel grid to be accessed is determined by accessing the pixel grids in the domain of the current accessed pixel grid, the number of the pixel grids to be accessed can be one or more; if one, the pixel grid to be accessed is written into the output queue; if more, the pixel grids to be accessed can be written into the output queue in the determined order of the pixel grids to be accessed.

[0146] S344, accessing the pixel grids in the domain of the current accessed pixel grid in the output queue to determine new pixel grids to be accessed.

[0147] Specifically, the pixel grids in the domain of the current accessed pixel grid in the output queue are accessed in the preset access order. The domain range of the current accessed pixel grid is not limited in the embodiments of the present application, for example, can be 4-neighbor domain or 8-neighbor domain. For example, as shown in FIG. 4, the pixel grids in the 8-neighbor domain of the current accessed pixel grid in the output queue are accessed. Figure 10

[0148] S345, marking the current accessed pixel grid in the output queue as an accessed pixel grid, and deleting the current accessed pixel grid in the process queue.

[0149] Specifically, since the current pixel grid in the output queue has been accessed at this time, it can be marked as an accessed pixel grid, and the current accessed pixel grid in the process queue can also be deleted from the process queue because it has been accessed.

[0150] S346, returning to execute the step of writing the pixel grid to be accessed into the process queue until there is no pixel grid to be accessed in the process queue.

[0151] Specifically, after determining the new pixel grid to be accessed, returning to execute step S343 to write the pixel grid to be accessed into the process queue, sequentially determining the current accessed pixel grid from the pixel grid to be accessed written into the process queue and writing into the output queue, and accessing the pixel grids in the domain of the current accessed pixel grid in the output queue. If there is a next new pixel grid to be accessed, continue to return to execute step S343 to write the pixel grid to be accessed into the process queue, so as to realize the access to all the pixel grids to be accessed in the connected domain of the selected pixel grid to be accessed from the pixel grid map. If there is no next new pixel grid to be accessed, it means that there is no pixel grid to be accessed in the connected domain of the selected pixel grid to be accessed from the pixel grid map. At this time, it can be considered that the connected domain corresponding to the selected pixel grid to be accessed from the pixel grid map is determined. ​

[0152] S347, return to execute the step of selecting one current access pixel grid from the pixel grid map until all the pixel grids to be accessed in the pixel grid map are accessed.

[0153] Specifically, after determining the connected domain of the selected one pixel grid to be accessed from the pixel grid map, return to step S341, select another pixel grid to be accessed from the pixel grid map, continue to execute steps S342 to S345, create a new output queue for storing the accessed pixel grids corresponding to the another pixel grid to be accessed, and finally the accessed pixel grids stored in the output queue can determine the connected domain of the another pixel grid to be accessed. In this way, until all the pixel grids to be accessed in the pixel grid map are accessed, i.e., all the connected domains in the pixel grid map are determined.

[0154] S348, determine the connected domain formed by the pixel grids written in each output queue as one parking line feature.

[0155] Specifically, the pixel grids written in each output queue form a connected domain, and each connected domain is one parking line feature.

[0156] Optionally, in step S350, according to the grid coordinates of each pixel grid in the pixel grid map, extract an outer contour pixel grid set of each parking line feature, including:

[0157] S351, for the connected domain corresponding to each parking line feature, select a starting pixel grid from the boundary of the connected domain, and take the starting pixel grid as a current pixel grid.

[0158] The starting pixel grid is the first grid selected from the boundary of the connected domain. The current pixel grid is the pixel grid to be accessed currently.

[0159] Specifically, the pixel grids contained in the outer contour grid set should be theoretically all the pixel grids on the boundary of the connected domain corresponding to the parking line feature. Therefore, a starting pixel grid is randomly selected from the boundary of the connected domain as the current pixel grid.

[0160] S352, based on a preset access sequence, access the pixel grids in the domain of the current pixel grid in turn.

[0161] The preset access sequence can be a clockwise sequence or a counterclockwise sequence, and the starting access pixel grid position when accessing the pixel grids in the domain can be set according to actual needs, for example, the pixel grid at the upper right corner in the domain of the current pixel grid is accessed each time; or the pixel grid closest to the last accessed current pixel grid in the domain is accessed.

[0162] Specifically, based on the starting access pixel grid position in the preset access order, the pixel grids in the domain of the current pixel grid are accessed, it is judged whether the pixel grid being accessed is the target pixel grid; if yes, step S353 is executed; if no, the pixel grid corresponding to the next pixel grid position is accessed according to the preset access order. If all the pixel grids in the domain of the current pixel grid are accessed, but there is no target pixel grid, it is indicated that step S355 is executed, and the search for the outer contour pixel grid is stopped.

[0163] S353, if the target pixel grid in the domain of the current pixel grid is accessed, and the grid coordinates of the target pixel grid and the starting pixel grid are different, the target pixel grid is added to the outer contour pixel grid set, and the target pixel grid is taken as the new current pixel grid.

[0164] The target pixel grid is a pixel grid with a grid score greater than a preset threshold and located in the pixel grid map. The preset threshold here can also be zero or a score threshold, so as to exclude the pixel grid as noise.

[0165] Specifically, if the pixel grid being accessed in the domain of the current pixel grid is the target pixel grid, it is judged whether the grid coordinates of the target pixel grid and the starting pixel grid are the same. If different, it indicates that the traversal of all the pixel grids in the connected domain corresponding to the current parking line element is not completed, at this time, the target pixel grid is added to the outer contour pixel grid set as the outer contour pixel grid, and the target pixel grid is taken as the new current pixel grid (i.e. the next current pixel grid), and step S354 is executed to access the pixel grids in the domain of the new current pixel grid. If the same, it indicates that the traversal of all the pixel grids in the connected domain corresponding to the current parking line element is completed, at this time, S355 is executed, and the search for the outer contour pixel grid is stopped.

[0166] S354, return to execute the step of sequentially accessing the pixel grids in the domain of the current pixel grid based on the preset access order.

[0167] Specifically, after the new current pixel grid is determined, the step of sequentially accessing the pixel grids in the domain of the current pixel grid based on the preset access order is returned to be executed, and the pixel grids in the domain of the new current pixel grid are accessed.

[0168] S355, until there is no target pixel grid in the domain of the current pixel grid or the grid coordinates of the target pixel grid and the starting pixel grid are the same, the access is stopped, and the outer contour pixel grid set of the parking line element is obtained.

[0169] Specifically, if the target pixel raster does not exist in the neighborhood of the current pixel raster, the process of finding the outer contour of the connected component stops; or if the raster coordinates of the target pixel raster and the starting pixel raster are the same, it means that the traversal of all pixel rasteres in the connected component corresponding to the current parking line feature is complete, and the access can be stopped. The outer contour pixel raster set of the parking line feature is obtained. The outer contour of the parking line feature composed of pixel rasteres can be determined based on the outer contour pixel raster in the outer contour pixel raster set.

[0170] Optional, the preset access order includes:

[0171] If the current pixel grid is the starting pixel grid, then select a pixel grid within the neighborhood of the starting pixel grid as the starting pixel grid to be accessed, and access the pixels in a counter-clockwise order.

[0172] If the current pixel grid is the target pixel grid, then the pixel grid closest to the previous current pixel grid within the target pixel grid's neighborhood is determined as the starting pixel grid to be visited, and the visits are performed in a counter-clockwise order.

[0173] Specifically, if the current pixel grid is the starting pixel grid, then select a pixel grid within the range of the starting pixel grid as the starting access pixel grid, and start accessing the pixel grid in a counter-clockwise order from the starting access pixel grid.

[0174] If the current pixel grid is the target pixel grid, then determine the distance between each pixel grid in the target pixel grid's neighborhood and the previous current pixel grid, and take the pixel grid corresponding to the minimum distance as the starting pixel grid to be visited, and visit the pixel grid in a counter-clockwise order starting from the starting pixel grid.

[0175] For example, such as Figure 10 As shown, if the current pixel grid is the target pixel grid B, and the previous current pixel grid is A (which is also pixel grid 4 within the 8-neighborhood of the current pixel grid B), then the pixel grid closest to the previous current pixel grid A within the 8-neighborhood of the target pixel grid B is determined to be either pixel grid 3 or pixel grid 5. Therefore, pixel grid 3 can be determined as the starting pixel grid to be visited, and the visits can proceed in a counter-clockwise order. Alternatively, pixel grid 5 can also be used as the starting pixel grid to be visited, and the visits can proceed in a counter-clockwise order.

[0176] Example 4

[0177] Figure 11 This is a schematic diagram of the structure of a garage semantic map determination device provided in Embodiment 4 of the present invention. Figure 11 As shown, the device includes: an information acquisition module 410, a semantic stitching module 420, a grid scoring module 430, and a semantic map determination module 440;

[0178] The information acquisition module 410 is configured to acquire vehicle motion information of a target vehicle moving along a planned path in a target garage and parking space line images collected by a camera arranged on the target vehicle.

[0179] The semantic splicing module 420 is configured to perform semantic splicing on the parking space line images according to the vehicle motion information of the target vehicle to obtain an initial semantic map of the target garage.

[0180] The grid scoring module 430 is configured to rasterize the initial semantic map to obtain a pixel grid map, score pixel grids in the pixel grid map, and determine grid scores of the pixel grids.

[0181] The semantic map determination module 440 is configured to extract a set of outer contour pixel grids of each parking space line element in the pixel grid map according to the grid scores, and determine a target semantic map according to the set of outer contour pixel grids.

[0182] Optionally, the semantic splicing module 420 comprises:

[0183] The coordinate creation unit is configured to create a collection time body coordinate system corresponding to each image collection time of the target vehicle in the moving process.

[0184] The first coordinate determination unit is configured to determine a parking space line pixel point on a parking space line image collected at each image collection time in the corresponding collection time body coordinate system.

[0185] The second coordinate determination unit is configured to determine, for each parking space line image, an initial body coordinate of the collection time body coordinate of the parking space line pixel point in an initial body coordinate system according to the vehicle motion information of the target vehicle, the initial body coordinate system being a body coordinate system corresponding to an initial motion position of the target vehicle.

[0186] The semantic splicing unit is configured to perform semantic splicing on the parking space line pixel points in the parking space line image in the initial body coordinate system to obtain the initial semantic map of the target garage.

[0187] Optionally, the first coordinate determination unit comprises:

[0188] The correction sub-unit is configured to perform distortion correction on the parking space line image collected at each image collection time.

[0189] The perception sub-unit is configured to perform pixel point perception on the distortion-corrected parking space line image to obtain a parking space line pixel point corresponding to each parking space line.

[0190] The first coordinate determination subunit is configured to determine, according to the intrinsic parameters and the extrinsic parameters of the camera, the acquisition time carrier coordinates of the parking space line pixel points in the corresponding acquisition time carrier coordinate system.

[0191] The extrinsic parameters of the camera include a transformation matrix of a camera coordinate system of the camera and a corresponding acquisition time carrier coordinate system.

[0192] Optionally, the grid scoring module 430 comprises:

[0193] The map generation unit is configured to rasterize the initial semantic map based on a preset size of a grid, to obtain a pixel grid map of the initial semantic map.

[0194] The third coordinate determination unit is configured to determine, according to the initial carrier coordinates of the parking space line pixel points in the initial semantic map, the pixel grid coordinates corresponding to each parking space line pixel point in the pixel grid map.

[0195] The number determination unit is configured to determine, according to the pixel grid coordinates corresponding to each parking space line pixel point, the number of parking space line pixel points contained in each pixel grid.

[0196] The score determination unit is configured to determine the grid score of the pixel grid according to the number of pixel points in the pixel grid.

[0197] Optionally, the grid scoring module 430 further comprises:

[0198] The zero setting unit is configured to set the grid score of the pixel grid to zero if the grid score of the pixel grid is less than a score threshold.

[0199] The score threshold is determined according to the number of pixel grids with a non-zero grid score in the pixel grid map, the preset size of the grid, and a perception error rate of the pixel points.

[0200] Optionally, the semantic map determination module 440 comprises:

[0201] The parking space line element determination unit is configured to determine, according to the grid score of each pixel grid in the pixel grid map, a parking space line element in the pixel grid map; the parking space line element is a connected domain composed of a plurality of pixel grids.

[0202] The set extraction unit is configured to extract, according to the grid coordinates of each pixel grid in the pixel grid map, an outer contour pixel grid set of each parking space line element.

[0203] The fourth coordinate determination unit is configured to determine the initial carrier coordinates of the pixel grid coordinates corresponding to the pixel grids in the outer contour pixel grid set in the initial carrier coordinate system.

[0204] a semantic map determination unit, configured to determine the target semantic map according to initial carrier coordinates of the pixel grids in each of the sets of outer contour pixel grids.

[0205] Optionally, the parking line element determination unit is specifically configured to:

[0206] selecting a pixel grid to be accessed from the pixel grid map; the pixel grid to be accessed is a pixel grid with a grid score greater than a preset threshold and not yet accessed;

[0207] creating an output queue and a process queue;

[0208] writing the pixel grid to be accessed into the process queue, and writing a current pixel grid determined from the pixel grid to be accessed in the process queue into the output queue;

[0209] accessing pixel grids in the domain of the current pixel grid in the output queue to determine new pixel grids to be accessed;

[0210] marking the current pixel grid in the output queue as an accessed pixel grid, and deleting the current pixel grid in the process queue;

[0211] returning to execute the step of writing the pixel grid to be accessed into the process queue until there is no pixel grid to be accessed in the process queue;

[0212] returning to execute the step of selecting a current pixel grid from the pixel grid map until all the pixel grids to be accessed in the pixel grid map are accessed;

[0213] determining a connected domain formed by the pixel grids written in each of the output queues as a parking line element.

[0214] Optionally, the set extraction unit is specifically configured to:

[0215] for each connected domain corresponding to a parking line element, selecting a starting pixel grid from the boundary of the connected domain, and taking the starting pixel grid as a current pixel grid;

[0216] sequentially accessing pixel grids in the domain of the current pixel grid based on a preset access order;

[0217] If a target pixel grid in the field of the current pixel grid is accessed, and the grid coordinates of the target pixel grid and the start pixel grid are different, the target pixel grid is added to the set of the outer contour pixel grids, and the target pixel grid is taken as a new current pixel grid; wherein the target pixel grid is a pixel grid with a grid score greater than a preset threshold and located in the pixel grid map;

[0218] The step of sequentially accessing the pixel grids in the field of the current pixel grid based on the preset access sequence is returned to be executed;

[0219] Until there is no target pixel grid in the field of the current pixel grid or the grid coordinates of the target pixel grid and the start pixel grid are the same, the access is stopped, and the set of the outer contour pixel grids of the parking space line element is obtained.

[0220] Optionally, the preset access sequence comprises:

[0221] If the current pixel grid is a start pixel grid, a pixel grid in the field of the start pixel grid is selected as a start access pixel grid, and the access is performed in a counterclockwise order;

[0222] If the current pixel grid is a target pixel grid, a pixel grid closest to the last current pixel grid in the field of the target pixel grid is determined as a start access pixel grid, and the access is performed in a counterclockwise order.

[0223] The device can execute the determination method of the garage semantic map provided by any embodiment of the application, and has corresponding function modules and beneficial effects of executing the determination method of the garage semantic map.

[0224] Embodiment five

[0225] Figure 12 A structural schematic diagram of an electronic device for implementing the determination method of the garage semantic map of the embodiment of the application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections, and relationships, and their functions, are merely examples and are not intended to limit the implementations described and / or claimed in this document.

[0226] As Figure 12As shown, the electronic device 10 includes at least one processor 11, and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., connected to the at least one processor 11 in communication. The memory stores a computer program executable by the at least one processor 11, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0227] Various components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc., an output unit 17, such as various types of displays, a speaker, etc., a storage unit 18, such as a magnetic disk, an optical disk, etc., and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0228] The processor 11 can be various general and / or special-purpose processing components having processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs various methods and processes described above, such as the determination method of the garage semantic map.

[0229] In some embodiments, the determination method of the garage semantic map can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the determination method of the garage semantic map described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the determination method of the garage semantic map by any other appropriate means, such as by means of firmware.

[0230] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a load programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0231] Computer programs used to implement the processes of the application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer program

[0232] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store computer programs for use by or in connection with an instruction execution system, apparatus, or device. Computer-readable storage media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium will include one or more lines of a program of instructions in a transitory signal, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0233] To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device 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 a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0234] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0235] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. A 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 defects of large management difficulty and weak business scalability in traditional physical host and VPS service.

[0236] It should be understood that the various forms of flow shown above can be re-ordered, added to, or deleted from without departing from the scope of the present disclosure. For example, the steps recited in the present disclosure can be executed in parallel, executed in sequence, or executed in a different order, as long as the desired results of the present disclosure are achieved, and the present disclosure is not limited herein.

[0237] The specific embodiments described above are not intended to be limiting, and persons skilled in the art will appreciate that various modifications, combinations, sub-combinations and alternatives can be made to the specific embodiments without departing from the spirit and principles of the disclosure. Accordingly, the disclosure is not limited to the specific embodiments described above, but only by the scope of the appended claims.

Claims

1. A method for determining a garage semantic map, characterized in that, The method comprises: acquiring vehicle motion information of a target vehicle moving along a planned path in a target garage and a parking space line image collected by a camera arranged on the target vehicle; performing parking space line semantic splicing on the parking space line image according to the vehicle motion information of the target vehicle to obtain an initial semantic map of the target garage; performing rasterization on the initial semantic map to obtain a pixel raster map, scoring pixel rasters in the pixel raster map, and determining raster scores of the pixel rasters; extracting a set of outer contour pixel rasters of each parking space line element in the pixel raster map according to the raster scores, and determining a target semantic map according to the set of outer contour pixel rasters. The rasterization of the initial semantic map to obtain the pixel raster map, the scoring of the pixel rasters in the pixel raster map, and the determination of the raster scores of the pixel rasters comprise: performing rasterization on the initial semantic map based on a preset size of a raster to obtain a pixel raster map of the initial semantic map; determining pixel raster coordinates corresponding to each parking space line pixel point in the pixel raster map according to initial carrier coordinates of the parking space line pixel points in the initial semantic map; determining the number of parking space line pixel points contained in each pixel raster according to the pixel raster coordinates corresponding to each parking space line pixel point; determining the raster scores of the pixel rasters according to the number of pixel points in the pixel rasters. The extraction of the set of outer contour pixel rasters of each parking space line element in the pixel raster map according to the raster scores and the determination of the target semantic map according to the set of outer contour pixel rasters comprise: determining parking space line elements in the pixel raster map according to the raster scores of each pixel raster in the pixel raster map; the parking space line elements are connected domains formed by a plurality of pixel rasters; extracting a set of outer contour pixel rasters of each parking space line element according to the raster coordinates of each pixel raster in the pixel raster map; determining initial carrier coordinates of the pixel raster coordinates corresponding to the pixel rasters in the set of outer contour pixel rasters in an initial carrier coordinate system; determining the target semantic map according to the initial carrier coordinates of the pixel rasters in each set of outer contour pixel rasters.

2. The method of claim 1, wherein, The performing of parking space line semantic splicing on the parking space line image according to the vehicle motion information of the target vehicle to obtain the initial semantic map of the target garage comprises: creating a carrier coordinate system corresponding to each image collection time of the target vehicle during movement; determining carrier coordinates of parking space line pixel points on the parking space line image collected at each image collection time in a carrier coordinate system corresponding to the image collection time; for each parking space line image, determining initial carrier coordinates of the carrier coordinates of the parking space line pixel points in a carrier coordinate system corresponding to the image collection time in an initial carrier coordinate system corresponding to an initial movement position of the target vehicle according to the vehicle motion information of the target vehicle; performing parking space line semantic splicing on the parking space line pixel points in the parking space line image in the initial carrier coordinate system to obtain the initial semantic map of the target garage.

3. The method of claim 2, wherein, Determine the parking space line pixel points on the parking space line image collected at each image collection time, and the collection time carrier coordinates in the corresponding collection time carrier coordinate system, including: Correct the distortion of the parking space line image collected at each image collection time; Perform pixel point perception on the distortion-corrected parking space line image to obtain the parking space line pixel points corresponding to each parking space line; Determine the collection time carrier coordinates of the parking space line pixel points in the corresponding collection time carrier coordinate system according to the intrinsic and extrinsic parameters of the camera; The extrinsic parameters of the camera include the transformation matrix of the camera coordinate system of the camera and the corresponding collection time carrier coordinate system.

4. The method of claim 1, wherein, Further comprising: If the grid score of the pixel grid is less than the score threshold, set the grid score of the pixel grid to zero; The score threshold is determined according to the number of pixel grids with a non-zero grid score in the pixel grid map, the preset size of the grid, and the perception error rate of the pixel point.

5. The method of claim 1, wherein, The parking space line elements in the pixel grid map are determined according to the grid score of each pixel grid in the pixel grid map, including: Select a to-be-visited pixel grid from the pixel grid map; the to-be-visited pixel grid is a pixel grid with a grid score greater than a preset threshold and not yet visited; Create an output queue and a process queue; Write the to-be-visited pixel grid into the process queue, and write the current visited pixel grid determined from the to-be-visited pixel grid in the process queue into the output queue; visit the pixel grids in the domain of the current visited pixel grid in the output queue to determine new to-be-visited pixel grids; Mark the current pixel grid in the output queue as a visited pixel grid, and delete the current visited pixel grid in the process queue; Return to execute the step of writing the to-be-visited pixel grid into the process queue until there is no to-be-visited pixel grid in the process queue; Return to execute the step of selecting a current visited pixel grid from the pixel grid map until all to-be-visited pixel grids in the pixel grid map are visited; Determine the connected domain formed by the pixel grid written in each output queue as a parking space line element.

6. The method of claim 1, wherein, The outer contour pixel grid set of each parking space line element is extracted according to the grid coordinates of each pixel grid in the pixel grid map, including: For each connected domain corresponding to a parking space line element, select a starting pixel grid from the boundary of the connected domain, and take the starting pixel grid as a current pixel grid; Access the pixel grids in the domain of the current pixel grid in turn based on a preset access order; If a target pixel grid in the domain of the current pixel grid is accessed, and the grid coordinates of the target pixel grid and the starting pixel grid are different, add the target pixel grid to the outer contour pixel grid set, and take the target pixel grid as a new current pixel grid; wherein the target pixel grid is a pixel grid with a grid score greater than a preset threshold and located in the pixel grid map; returning to performing the step of accessing the pixel grids in the domain of the current pixel grid in turn according to the preset access sequence; stopping the access until there is no target pixel grid in the domain of the current pixel grid or the grid coordinates of the target pixel grid and the starting pixel grid are the same, and obtaining the set of outer contour pixel grids of the parking line element.

7. The method of claim 6, wherein, The preset access sequence comprises: if the current pixel grid is a starting pixel grid, selecting a pixel grid from the domain of the starting pixel grid as a starting access pixel grid, and performing the access in a counterclockwise order; if the current pixel grid is a target pixel grid, determining the pixel grid closest to the last current pixel grid from the domain of the target pixel grid as a starting access pixel grid, and performing the access in a counterclockwise order.

8. A device for determining a garage semantic map, characterized in that, comprise: an information acquisition module, configured to acquire vehicle motion information of a target vehicle when the target vehicle moves along a planned path in a target garage and a parking line image collected by a camera arranged on the target vehicle; a semantic splicing module, configured to perform semantic splicing of the parking line image according to the vehicle motion information of the target vehicle to obtain an initial semantic map of the target garage; a grid scoring module, configured to rasterize the initial semantic map to obtain a pixel grid map, score the pixel grids in the pixel grid map, and determine grid scores of the pixel grids; a semantic map determination module, configured to extract a set of outer contour pixel grids of each parking line element in the pixel grid map according to the grid scores, and determine a target semantic map according to the set of outer contour pixel grids. The grid scoring module comprises: a map generation unit, configured to rasterize the initial semantic map based on a preset size of a grid to obtain a pixel grid map of the initial semantic map; a third coordinate determination unit, configured to determine pixel grid coordinates corresponding to each parking line pixel point in the pixel grid map according to initial carrier coordinates of the parking line pixel points in the initial semantic map; a quantity determination unit, configured to determine a number of parking line pixel points included in each pixel grid according to the pixel grid coordinates corresponding to each parking line pixel point; a score determination unit, configured to determine the grid score of the pixel grid according to the number of pixel points in the pixel grid. The semantic map determination module comprises: a parking line element determination unit, configured to determine a parking line element in the pixel grid map according to the grid score of each pixel grid in the pixel grid map; the parking line element is a connected domain composed of multiple pixel grids; a set extraction unit, configured to extract a set of outer contour pixel grids of each parking line element according to the grid coordinates of each pixel grid in the pixel grid map; a fourth coordinate determination unit, configured to determine initial carrier coordinates of the pixel grid coordinates corresponding to the pixel grids in the set of outer contour pixel grids in an initial carrier coordinate system; a semantic map determination unit, configured to determine the target semantic map according to the initial carrier coordinates of the pixel grids in each set of outer contour pixel grids.

9. An electronic device, comprising: The electronic device comprises: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the method for determining a garage semantic map according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for enabling the processor to implement the method for determining a garage semantic map according to any one of claims 1-7 when executed.

Images