Parking space corner point recognition method, storage medium, electronic device, and vehicle

Abstract

The application provides a parking space corner point recognition method, a storage medium, an electronic device and a vehicle. The method comprises the following steps: acquiring a parking vehicle corner point coordinate in a panoramic image of a parking vehicle; determining a target area of a parking space according to the parking vehicle corner point coordinate and a parking vehicle coordinate in the panoramic image; and when a target parking vehicle corner point is missing according to the target area, performing corner point restoration on a missing area of the target parking vehicle corner point. The application avoids poor recognition accuracy caused by parking space corner point occlusion, avoids incorrect parking poses, and thus provides support for improving the accuracy of vehicle parking.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicles, in particular to a parking space corner point recognition method, a storage medium, an electronic device and a vehicle. BACKGROUND

[0002] In the field of automatic driving, images are an important means for vehicles to obtain information about the outside world. However, in the process of automatic parking of a vehicle, the recognition accuracy may be inaccurate, which may result in incorrect parking pose and affect the user's parking experience. SUMMARY

[0003] The present application aims to at least solve one of the technical problems existing in the prior art.

[0004] To this end, one object of the present application is to provide a parking space corner point recognition method which avoids poor recognition accuracy caused by parking space corner point occlusion and avoids incorrect parking pose, thereby providing support for improving the accuracy of vehicle parking.

[0005] To this end, a second object of the present application is to provide a storage medium.

[0006] To this end, a third object of the present application is to provide an electronic device.

[0007] To this end, a fourth object of the present application is to provide a vehicle.

[0008] To achieve the above objects, an embodiment of the first aspect of the present application provides a parking space corner point recognition method, comprising: obtaining a parking vehicle corner point coordinate of a parking vehicle in a panoramic image of the parking vehicle; determining a target region of a parking space according to the parking vehicle corner point coordinate and a parking vehicle coordinate in the panoramic image; and performing corner point restoration on a missing region of a target parking space corner point when the target parking space corner point is missing according to the target region.

[0009] According to the parking space corner point recognition method of the present application, the target region of the parking space is determined by combining the parking vehicle corner point coordinate and the parking vehicle coordinate, the target region is processed to obtain the target parking space corner point, i.e. the real parking space corner point, and when the parking space feature such as the parking space corner point is occluded or exceeds the line of sight, etc., the effective feature cannot be formed, the missing region of the target vehicle corner point is restored to obtain the effective parking space corner point feature, thereby assisting the parking of the parking vehicle according to the effective parking space corner point feature. By restoring the missing region of the target parking space corner point, poor recognition accuracy caused by parking space corner point occlusion can be avoided, and incorrect parking pose can be avoided, thereby providing support for improving the accuracy of vehicle parking.

[0010] In some embodiments, determining the target region of the parking space according to the parking space corner point coordinates and the parking vehicle coordinates in the panoramic image comprises: determining a near corner point coordinate in the parking space corner point coordinates; determining a region of interest of the parking space according to the near corner point coordinate and the parking vehicle coordinates, and taking the region of interest as the target region of the parking space.

[0011] In some embodiments, determining the target parking space corner point according to the target region comprises: performing edge extraction on the region of interest to obtain an edge line of the region of interest; determining a direction vector of the edge line according to the edge line; and determining the target parking space corner point according to the direction vector and a preset feature description vector.

[0012] In some embodiments, performing edge extraction on the region of interest comprises: performing edge extraction on the region of interest in an RGB channel.

[0013] In some embodiments, determining the direction vector of the edge line according to the edge line comprises: performing a preset transformation on the edge line to obtain the direction vector of the edge line.

[0014] In some embodiments, determining the target parking space corner point according to the direction vector and the preset feature description vector comprises: screening the direction vector according to an included angle of a preset parking space mark line; determining a set of intersection points formed by the direction vector according to the screened direction vector, and taking the set of intersection points as a to-be-determined set of the parking space corner points; and performing feature matching on the to-be-determined set of the parking space corner points according to a preset feature description vector to obtain the target parking space corner point.

[0015] In some embodiments, performing feature matching on the to-be-determined set of the parking space corner points according to the preset feature vector to obtain the target parking space corner point comprises: calculating a dot product of the preset feature description vector and a description vector in the to-be-determined set of the parking space corner points; and taking an intersection point in the to-be-determined set whose dot product satisfies a preset condition as the target parking space corner point.

[0016] In some embodiments, determining the target region of the parking space according to the parking space corner point coordinates and the parking vehicle coordinates in the panoramic image comprises: determining a parking space corner point coordinate in a preset confidence region, wherein the preset confidence region is a low confidence region; and performing cropping on the panoramic image according to the parking space corner point coordinate in the preset confidence region and the parking vehicle coordinates to obtain the target region.

[0017] In some embodiments, determining the parking space corner point coordinate in the preset confidence region comprises: if the parking space corner point coordinate is in a set region range, determining that the parking space corner point coordinate is in the preset confidence region.

[0018] In some embodiments, after the panoramic image is cropped according to the parking space corner point coordinates of the preset area and the parking vehicle coordinates to obtain the target area, the method further comprises: if the target area intersects with the area where the vehicle is located, adjusting the target area until the target area does not intersect with the area where the vehicle is located.

[0019] In some embodiments, determining that the target parking space corner point is missing comprises: judging whether the target parking space corner point is blocked; if yes, determining that the target parking space corner point is missing.

[0020] In some embodiments, restoring the missing area of the target parking space corner point comprises: obtaining the coordinates of a parking space corner point adjacent to the target parking space corner point; determining geometric proportion data of the adjacent parking space corner point coordinates and the target parking space corner point; and completing the missing of the target parking space corner point according to the geometric proportion data.

[0021] To achieve the above object, the second aspect of the embodiments of the present application provides a storage medium, wherein the storage medium stores a parking space corner point identification program, and the parking space corner point identification program is executed by a processor to implement the parking space corner point identification method as described in the above embodiments.

[0022] According to the storage medium of the embodiments of the present application, the target area of the parking space is determined by combining the parking space corner point coordinates of the parking vehicle and the parking vehicle coordinates, the target area is processed to obtain the target parking space corner point, i.e., the real parking space corner point, and when the parking space feature such as the parking space corner point is blocked or exceeds the line of sight, etc., the missing area of the target vehicle corner point is restored to obtain the effective parking space corner point feature, so that the parking vehicle is parked according to the effective parking space corner point feature, and the missing area of the target parking space corner point is restored, which can avoid the poor recognition accuracy caused by the blocking of the parking space corner point and avoid the incorrect parking pose, thereby providing support for improving the accuracy of vehicle parking.

[0023] To achieve the above object, the third aspect of the embodiments of the present application provides an electronic device, which comprises: a processor, a memory, and a parking space corner point identification program stored in the memory and executable on the processor, and the parking space corner point identification program is executed by the processor to implement the parking space corner point identification method as described in the above embodiments.

[0024] According to the electronic device of the embodiment of the present application, the target area of the parking space is determined by combining the parking space angle point coordinates of the parked vehicle and the parked vehicle coordinates, the target area is processed, the target parking space angle point, that is, the real parking space angle point, is obtained, and when the parking space features such as the parking space angle point are blocked or exceed the range of the line of sight, the effective parking space angle point features are obtained by restoring the angle points of the missing area of the target vehicle angle point, so that the parked vehicle is parked according to the effective parking space angle point features. By restoring the angle points of the missing area of the target parking space angle point, the poor recognition accuracy caused by the parking space angle point blocking can be avoided, and the parking pose can be corrected, thereby providing support for improving the accuracy of vehicle parking.

[0025] In order to achieve the above-mentioned purpose, the embodiment of the fourth aspect of the present application proposes a vehicle, which comprises the electronic device described in the above-mentioned embodiments.

[0026] According to the vehicle of the embodiment of the present application, the target area of the parking space is determined by combining the parking space angle point coordinates of the parked vehicle and the parked vehicle coordinates, the target area is processed, the target parking space angle point, that is, the real parking space angle point, is obtained, and when the parking space features such as the parking space angle point are blocked or exceed the range of the line of sight, the effective parking space angle point features are obtained by restoring the angle points of the missing area of the target vehicle angle point, so that the parked vehicle is parked according to the effective parking space angle point features. By restoring the angle points of the missing area of the target parking space angle point, the poor recognition accuracy caused by the parking space angle point blocking can be avoided, and the parking pose can be corrected, thereby providing support for improving the accuracy of vehicle parking.

[0027] Additional aspects and advantages of the present application will be made apparent from the following description of the embodiments of the present application, which will be made with reference to the accompanying drawings, from which the aspects and advantages can become readily apparent, or can be learned by the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0028] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

[0029] Figure 1 is a schematic diagram of the existing deep learning target frame-based parking space detection technology according to an embodiment of the present application;

[0030] Figure 2 is a schematic diagram of the deep learning detection frame when the range of the line of sight is exceeded according to an embodiment of the present application;

[0031] Figure 3 is a flowchart of a parking space angle point recognition method according to an embodiment of the present application;

[0032] Figure 4is a schematic diagram of acquiring a region of interest according to an embodiment of the present application;

[0033] Figure 5 is a schematic diagram of corner point detection according to an embodiment of the present application;

[0034] Figure 6 is a flowchart of a method of identifying a parking space corner point according to an embodiment of the present application;

[0035] Figure 7 is a structural block diagram of an electronic device according to an embodiment of the present application;

[0036] Figure 8 is a structural block diagram of a vehicle according to an embodiment of the present application.

[0037] Reference signs: electronic device 2; processor 100; memory 101; parking space corner point identification program 102; vehicle 3. DETAILED DESCRIPTION

[0038] Embodiments of the present application are described in detail below, and the embodiments described with reference to the accompanying drawings are exemplary, and embodiments of the present application are described in detail below.

[0039] In the related art, as shown in Figure 1 , it is a schematic diagram of an existing parking space detection technology based on a deep learning target frame. The Autonomous Parking Assistance (APA) function usually adopts a deep learning model, uses a deep learning check frame to identify the corner points or markings formed by the markings on the parking space on a panoramic image, to detect the parking space. Since the deep learning model is based on the features formed by the parking space on the image to complete the parking space identification process, when the parking space features cannot form effective features due to reasons such as occlusion, exceeding the range of sight, etc., for example Figure 2 , as shown in the figure, part of the parking space features exceeds the range of sight, resulting in poor accuracy of parking space recognition, causing the parking pose to be incorrect, thereby affecting the user's parking experience.

[0040] Therefore, the method of identifying a parking space corner point according to an embodiment of the present application determines a target region of the parking space by combining the parking space corner point coordinates of the parking vehicle and the coordinates of the parking vehicle, determines the target region of the parking space, processes the target region to obtain a target parking space corner point, when the parking space features cannot form effective features due to reasons such as occlusion, exceeding the range of sight, etc., restores the corner points in the missing region of the target vehicle corner point to obtain effective parking space corner point features, thereby providing parking assistance to the parking vehicle, and improving the accuracy of parking the vehicle.

[0041] The method of identifying a parking space corner point according to an embodiment of the present application is described below in conjunction with Figures 3-6 .

[0042] As Figure 3 shown, the parking space corner point recognition method of the embodiment of the application at least includes steps S1-S3.

[0043] Step S1, obtain the parking space corner point coordinates of the parking vehicle in the panoramic image of the parking vehicle.

[0044] The panoramic image is an image of the surrounding environment during the parking process of the vehicle, and the panoramic image of the parking vehicle is obtained for the deep learning model to analyze the panoramic image.

[0045] The parking space corner point coordinates are coordinate information corresponding to the corner points formed by the parking space markings, and by determining the parking space corner point coordinates of the parking vehicle in the panoramic image, the specific position of the parking vehicle that can be parked is determined, so as to improve the accuracy of the parking vehicle entering the target parking space.

[0046] In an embodiment, during the driving of the parking vehicle to the target parking space, the camera is used to collect the image of the surrounding environment of the vehicle in real time, the image is subjected to projective transformation, and the image overlap part is spliced to convert into a 360° panoramic image for information extraction by the multi-task deep learning model.

[0047] After obtaining the panoramic image of the parking vehicle, the deep learning module processes the panoramic image, such as information extraction, to obtain four parking space corner points of the parking space that can be parked, such as (P0, P1, P2, and P3), wherein P0 and P1 are near corner points and P2 and P3 are far corner points. After determining the parking space corner points of the parking space, the corresponding parking space corner point coordinates are determined, such as the parking space corner point coordinates of P0 (U0, V0), the parking space corner point coordinates of P1 (U1, V1), the parking space corner point coordinates of P2 (U2, V2), and the parking space corner point coordinates of P3 (U3, V3). That is, the parking space corner point coordinates are (U0, V0, U1, V1, U2, V2, U3, V3). By determining the parking space corner point coordinates of the parking vehicle, the accuracy of the vehicle parking is improved.

[0048] It should be noted that the upper left corner of the panoramic image is taken as the coordinate origin, the image width direction is the U direction, the image height direction is the V direction, and the order of the parking space corner point coordinates is counterclockwise.

[0049] Step S2, determine the target area of the parking space according to the parking space corner point coordinates and the parking vehicle coordinates in the panoramic image.

[0050] The parking vehicle coordinate in the panoramic image is real-time changing coordinate information during the parking process of the vehicle, and the target area of the parking space is the surrounding environment of the target parking space into which the parking vehicle needs to enter. By determining the parking vehicle coordinate in the panoramic image, the specific position of the parking vehicle can be determined. It can be understood that the parking vehicle coordinate changes according to different vehicle motion states.

[0051] In an embodiment, after the panoramic image in the preset range of the parking vehicle is acquired, the deep learning module processes the panoramic image, for example, performs information extraction, to acquire the parking vehicle coordinate, for example, denoted as (U car_min ,V car_min ,U car_max ,V car_max ), so as to determine the correct parking vehicle in the panoramic image and the corresponding position information of the parking vehicle in the panoramic image. In combination with the parking space corner point coordinates (U0, V0, U1, V1, U2, V2, U3, V3) of the parking vehicle in the panoramic image, the target area of the parking space is determined. By determining the parking vehicle coordinate in the panoramic image, the starting position of the parking vehicle is determined. By determining the parking space corner point coordinates, the target position of the parking vehicle is determined, so that the area from the starting position to the target position of the parking vehicle is determined as the target area.

[0052] In step S3, when the target parking space corner point is determined to be missing according to the target area, the missing area of the target parking space corner point is restored.

[0053] The target parking space corner point is specific position information of the target parking space into which the parking vehicle needs to enter. By determining the target parking space corner point according to the target area, the real parking space into which the parking vehicle enters can be determined.

[0054] In an embodiment, after the target area is determined, the target area is processed, the target area is mapped to the panoramic image, the target parking space corner point, that is, the real parking space corner point, is determined in the panoramic image, and the target parking space of the parking vehicle is determined.

[0055] After the target parking space corner point is determined, it is judged whether the target parking space corner point is missing. When the target parking space corner point is missing, it is considered that the vehicle may have poor recognition effect on the parking space due to factors such as sight distance, occlusion, and illumination. Then, the missing area of the target parking space corner point is restored to obtain effective parking space corner point features. According to the effective parking space corner point features, parking assistance is performed on the parking vehicle, so that the parking vehicle has good robustness, the recognition accuracy of the parking space is improved, and the parking pose of the parking vehicle is continuously corrected.

[0056] According to the parking space corner point recognition method in the embodiment of the present application, the target region of the parking space is determined by combining the parking space corner point coordinates and the parking vehicle coordinates, the target region is processed, the target parking space corner point, that is, the real parking space corner point is obtained, and when the parking space features such as the parking space corner point are blocked or beyond the line of sight and cannot form effective features, the missing region of the target vehicle corner point is restored, and the effective parking space corner point features are obtained, so that the parking vehicle is parked according to the effective parking space corner point features. By restoring the missing region of the target parking space corner point, the poor recognition accuracy caused by the parking space corner point blocking can be avoided, and the parking posture can be corrected, thereby providing support for improving the accuracy of vehicle parking.

[0057] In some embodiments, the target region of the parking space is determined according to the parking space corner point coordinates and the parking vehicle coordinates in the panoramic image, including: determining the near corner point coordinates in the parking space corner point coordinates; determining the region of interest of the parking space according to the near corner point coordinates and the parking vehicle coordinates, and taking the region of interest as the target region of the parking space.

[0058] In an embodiment, as shown in Figure 4 , it is a schematic diagram of obtaining the region of interest of an embodiment of the present application. After determining the parking space corner point coordinates of the parking space in the panoramic image, the parking space corner point coordinates of the near corner point P0 in the parking space corner point coordinates are determined as (U0, V0), and the parking space corner point coordinates of P1 are determined as (U1, V1). The region of interest of the parking space is determined according to the near corner point coordinates (U0, V0, U1, V1) and the parking vehicle coordinates (U car_min ,V car_min ,U car_max ,V car_max ), so as to maximize the effective features of the region of interest, take the region of interest as the target region of the parking space, exclude the interference of redundant information, and improve the detection effect of the parking space corner point.

[0059] In some embodiments, the target parking space corner point is determined according to the target region, including: performing edge extraction on the region of interest to obtain the edge lines of the region of interest; determining the direction vector of the edge lines according to the edge lines; and determining the target parking space corner point according to the direction vector and the preset feature description vector.

[0060] In an embodiment, as shown in Figure 5 , it is a schematic diagram of corner point detection of an embodiment of the present application. After obtaining the region of interest, the edge extraction is performed on the region of interest to obtain the edge lines of the region of interest. After obtaining the edge lines of the region of interest, the direction vector of the edge lines is determined by changing the edge lines, and the target parking space corner point is determined according to the direction vector satisfying the preset feature description vector.

[0061] In some embodiments, the edge extraction on the region of interest comprises: performing edge extraction on the region of interest in RGB channels.

[0062] In an embodiment, after the region of interest is obtained, edge extraction is performed on the region of interest in RGB three channels respectively by using Sobel, Prewitt, Roberts, Canny, Marr-Hildreth and the like, and the edges extracted in the three channels are superimposed to obtain all features of the parking space marker line in the color space, so as to enhance the edge features.

[0063] In some embodiments, the direction vector of the edge line is determined according to the edge line, comprising: performing a preset transformation on the edge line to obtain the direction vector of the edge line.

[0064] In an embodiment, as shown in Figure 5 , after the edge line of the region of interest is obtained, a preset transformation is performed on the edge line, for example, Hough transformation, to convert the edge line into a vector form to obtain the direction vector of the edge line, so as to retain the semantic information of the parking space marker line.

[0065] In some embodiments, the target parking corner point is determined according to the direction vector and the preset feature descriptor, comprising: screening the direction vector according to the included angle of the preset parking space marker line; determining an intersection set formed by the screened direction vector as a to-be-determined set of parking corner points; and performing feature matching on the to-be-determined set of parking corner points according to the preset feature descriptor to obtain the target parking corner point.

[0066] In an embodiment, as shown in Figure 5 , after the direction vector of the edge line is obtained, the direction vector is screened according to the included angle of the preset parking space marker line, for example, 90°, the screened direction vector satisfies the included angle of the preset parking space marker line, so as to retain the direction vector with high similarity to the horizontal and vertical directions of the target parking space as a possible parking space boundary line; the intersection set formed by the screened direction vector is calculated, and the intersection points located outside the region of interest are excluded, the intersection set is determined as the to-be-determined set of parking corner points, and the feature matching is performed between the preset feature descriptor and the to-be-determined set of parking corner points, so as to screen out the real parking corner point from the to-be-determined set to obtain the target parking corner point, so that the vehicle has good robustness, and is thus widely applicable to parking space recognition under various working conditions.

[0067] In some embodiments, the feature matching is performed on the to-be-determined set of parking corner points according to the preset feature vector to obtain the target parking corner point, comprising: calculating the dot product of the preset feature descriptor and the descriptor in the to-be-determined set of parking corner points; and taking the intersection point of the to-be-determined set of descriptors satisfying the preset condition as the target parking corner point.

[0068] In the embodiments, after the to-be-determined set of parking corner points is determined, the to-be-determined set of parking corner points is traversed, a size relationship of pixel gray values in a range of three pixel units around the parking corner point is constructed according to the existing parking corner point, a feature description vector and a principal direction of the parking corner point are constructed, that is, a description vector in the to-be-determined set of parking corner points, a dot product of a preset feature description vector and the description vector in the to-be-determined set of parking corner points is calculated, and an intersection point in the to-be-determined set of parking corner points that satisfies a preset condition is taken as the target parking corner point. If the dot product of the description vector in the to-be-determined set of parking corner points does not satisfy the preset condition, it is considered that there is a special state in the region of interest, for example, the parking mark line has a state of blur, occlusion, etc., it is considered that the effective features of the region of interest are less, and the target parking corner point cannot be determined. The dot product of the description vector in the to-be-determined set of parking corner points is determined, so that the vehicle has good robustness, and thus the vehicle is widely applicable to parking space recognition under various working conditions.

[0069] In some embodiments, the target region of the parking space is determined according to the parking corner point coordinates and the parking vehicle coordinates in the panoramic image, including: determining the parking corner point coordinates in the preset confidence region, wherein the preset confidence region is a low confidence region; and cropping the panoramic image according to the parking corner point coordinates in the preset confidence region and the parking vehicle coordinates to obtain the target region.

[0070] In the embodiments, after the parking corner point coordinates of the parking space in the panoramic image are determined, it is determined whether the parking corner point coordinates are in the preset confidence region, that is, the low confidence region, and the panoramic image is dynamically cropped according to the parking corner point coordinates in the preset confidence region and the parking vehicle coordinates (U car_min ,V car_min ,U car_max ,V car_max ) to form a plurality of cropped images containing the parking mark line, the region of interest of the parking space is determined on the cropped images, the effective features of the region of interest are maximally reserved, the accuracy of subsequent mark line detection is improved, the amount of calculation of subsequent modules is reduced, the region of interest is taken as the target region of the parking space, the interference of redundant information irrelevant to the parking corner point is excluded, and the detection effect of the parking corner point is improved.

[0071] In some embodiments, the parking corner point coordinates in the preset confidence region are determined, including: if the parking corner point coordinates are in the set region range, it is determined that the parking corner point coordinates are in the preset confidence region.

[0072] In an embodiment, after the parking space corner point coordinates in the panoramic image are determined, it is determined whether the parking space corner point coordinates are in a set region range, for example, at a splicing seam of the panoramic image, far from the vehicle, or partially blocked by the vehicle itself. If the parking space corner point coordinates are in the set region range, it is determined that the parking space corner point coordinates are in a preset confidence region, i.e., a low-confidence region. If the parking space corner point coordinates are not in the set region range, it is determined that the parking space corner point coordinates are not in the preset confidence region, i.e., in a high-confidence region, and the high-confidence parking space corner point coordinates are output.

[0073] In some embodiments, after the panoramic image is cropped according to the parking space corner point coordinates in the preset confidence region and the parking vehicle coordinates to obtain a target region, the method further includes: if the target region intersects with a region where the vehicle is located, the target region is adjusted until the target region does not intersect with the region where the vehicle is located.

[0074] In an embodiment, after the panoramic image is cropped according to the parking space corner point coordinates in the preset confidence region and the parking vehicle coordinates to obtain a target region, it is determined whether the target region intersects with a region where the vehicle is located. If the target region intersects with the region where the vehicle is located, the target region is adjusted, i.e., the length of the edge in the U direction or the V direction of the target region is reduced, and the length of the edge in the V direction or the U direction is correspondingly increased, until the target region does not intersect with the region where the vehicle is located.

[0075] In some embodiments, after the target parking space corner point is determined according to the target region, the method further includes: if the target parking space corner point is missing, the missing region of the target parking space corner point is completed.

[0076] In an embodiment, after the target parking space corner point is determined according to the target region, it is determined whether the target parking space corner point is missing. If it is determined that the target parking space corner point is missing, the target parking space corner point recovery module completes the missing region of the target parking space corner point, thereby further improving the robustness of the parking space recognition method.

[0077] In some embodiments, determining that the target parking space corner point is missing includes: determining whether the target parking space corner point is blocked; if yes, it is determined that the target parking space corner point is missing.

[0078] In an embodiment, after the target parking space corner point is obtained, it is determined whether the target parking space corner point is blocked. If the target parking space corner point is blocked, it is determined that the target parking space corner point is missing, and the target parking space corner point recovery module needs to complete the missing region of the target parking space corner point. If the target parking space corner point is not blocked, the target parking space corner point is considered as a high-confidence parking space corner point coordinate, and the high-confidence parking space corner point coordinate is output.

[0079] In some embodiments, corner point restoration of the missing area of ​​the target parking space corner point includes: obtaining the coordinates of parking space corner points adjacent to the target parking space corner point; determining the geometric ratio data between the coordinates of the adjacent parking space corner points and the target parking space corner point; and filling in the missing corner point of the target parking space based on the geometric ratio data.

[0080] In this embodiment, when it is determined that the corner point of the target parking space is missing, the parking space corner point recovery module obtains the coordinates of the corner points of the parking spaces adjacent to the corner point of the target parking space, determines the geometric ratio data between the coordinates of the adjacent parking space corner points and the corner point of the target parking space, such as parallel, equidistant, perpendicular, etc., and restores the corner point in the missing area of ​​the target parking space corner point based on the geometric ratio data, thereby further improving the robustness of the parking space recognition method.

[0081] The following is for reference. Figure 6 The method for identifying parking space corner points according to an embodiment of the present invention will be illustrated by example.

[0082] like Figure 6 As shown, the parking space corner point identification method of this embodiment includes at least steps S12-S26.

[0083] Step S12: Obtain the coordinates of the parking space corner point of the parked vehicle in the panoramic image of the parked vehicle.

[0084] Step S13: Determine whether the coordinates of the corner point of the parking space are within the set area. If yes, proceed to step S14; otherwise, proceed to step S15.

[0085] Step S14: If the coordinates of the parking space corner point are determined to be outside the preset confidence range, then output the coordinates of the parking space corner point with high confidence.

[0086] Step S15: Determine that the corner coordinates of the parking space are within a preset confidence region. Based on the corner coordinates of the parking space in the preset confidence region and the vehicle coordinates, crop the panoramic image to obtain the target area.

[0087] Step S16: Determine whether there is an intersection between the target area and the area where the vehicle is located. If yes, proceed to step S17; otherwise, proceed to step S18.

[0088] Step S17: Adjust the target area until the target area and the area where the vehicle is located no longer overlap.

[0089] Step S18: Determine the coordinates of the nearest corner point in the corner coordinates of the parking space, and determine the region of interest of the parking space based on the coordinates of the near corner point and the coordinates of the parked vehicle.

[0090] Step S19: Extract the edge of the region of interest in the RGB channel to obtain the edge lines of the region of interest.

[0091] Step S20, preset transformation is performed on the edge lines to obtain a direction vector of the edge lines.

[0092] Step S21, the direction vector is screened according to a preset angle of the parking space mark line; and a set of intersection points formed by the screened direction vector is determined as a set of to-be-determined parking space corner points.

[0093] Step S22, a dot product of a preset feature description vector and a description vector in the set of to-be-determined parking space corner points is calculated, and an intersection point in the set of to-be-determined parking space corner points that satisfies a preset condition is taken as a target parking space corner point.

[0094] Step S23, it is judged whether the target parking space corner point is blocked, if yes, step S25 is performed; otherwise, step S24 is performed.

[0095] Step S24, a high-confidence parking space corner point coordinate is output.

[0096] Step S25, it is determined that the target parking space corner point is missing.

[0097] Step S26, a parking space corner point coordinate adjacent to the target parking space corner point is obtained, geometric proportion data of the adjacent parking space corner point coordinate and the target parking space corner point is determined, and the missing of the target parking space corner point is completed according to the geometric proportion data.

[0098] According to the parking space corner point recognition method provided in the embodiments of the present application, the target area of the parking space is determined by combining the parking space corner point coordinate of the parking vehicle and the parking vehicle coordinate, the target area is processed to obtain the target parking space corner point, i.e., the real parking space corner point, and when the parking space feature, such as the parking space corner point, is blocked or exceeds the line of sight, etc., the missing area of the target vehicle corner point is restored to obtain the effective parking space corner point feature, so that the parking vehicle is assisted in parking according to the effective parking space corner point feature. By restoring the corner point of the missing area of the target parking space corner point, the poor recognition accuracy caused by the blocking of the parking space corner point can be avoided, and the parking pose can be corrected, thereby providing support for improving the accuracy of vehicle parking.

[0099] The storage medium of the embodiments of the present application is described below.

[0100] The storage medium of the embodiments of the present application stores a parking space corner point recognition program, and the parking space corner point recognition program is executed by a processor to realize the parking space corner point recognition method described in any one of the embodiments of the present application.

[0101] The storage medium according to the embodiment of the present application determines the target area of the parking space by combining the parking space corner point coordinates of the parking vehicle and the parking vehicle coordinates, processes the target area, obtains the target parking space corner point, that is, the real parking space corner point, and when the parking space features, such as the parking space corner point, are blocked or beyond the line of sight, and effective features cannot be formed, restores the corner point of the missing area of the target vehicle corner point to obtain effective parking space corner point features, thereby providing parking assistance to the parking vehicle according to the effective parking space corner point features. By restoring the corner point of the missing area of the target parking space corner point, the poor recognition accuracy caused by the parking space corner point blocking can be avoided, and the parking pose can be corrected, thereby providing support for improving the accuracy of vehicle parking.

[0102] Reference will be made to Figure 7 The electronic device 2 according to the embodiment of the present application is described below.

[0103] As Figure 7 shown, the electronic device 2 according to the embodiment of the present application includes a processor 100, a memory 101, and a parking space corner point identification program 102 stored in the memory 101 and executable on the processor 100. When the parking space corner point identification program 102 is executed by the processor 100, the parking space corner point identification method described in any one of the above embodiments of the present application is implemented.

[0104] The electronic device 2 according to the embodiment of the present application determines the target area of the parking space by combining the parking space corner point coordinates of the parking vehicle and the parking vehicle coordinates, processes the target area, obtains the target parking space corner point, that is, the real parking space corner point, and when the parking space features, such as the parking space corner point, are blocked or beyond the line of sight, and effective features cannot be formed, restores the corner point of the missing area of the target vehicle corner point to obtain effective parking space corner point features, thereby providing parking assistance to the parking vehicle according to the effective parking space corner point features. By restoring the corner point of the missing area of the target parking space corner point, the poor recognition accuracy caused by the parking space corner point blocking can be avoided, and the parking pose can be corrected, thereby providing support for improving the accuracy of vehicle parking.

[0105] Reference will be made to Figure 8 The vehicle 3 according to the embodiment of the present application is described below.

[0106] As Figure 8 shown, the vehicle 3 according to the embodiment of the present application includes the electronic device 2 described in the above embodiments.

[0107] According to the vehicle 3 of the embodiment of the present application, the target area of the parking space is determined by combining the parking vehicle coordinate and the parking space angle point coordinate of the parking vehicle, the target area is processed, the target parking space angle point, i.e., the real parking space angle point, is obtained, and when the parking space features, such as the parking space angle point, are blocked or beyond the line of sight, the missing area of the target vehicle angle point is restored, the effective parking space angle point feature is obtained, and the parking assistance is provided for the parking vehicle according to the effective parking space angle point feature. The missing area of the target parking space angle point is restored, the poor recognition accuracy caused by the parking space angle point blocking is avoided, the parking pose is not incorrect, and the support is provided for improving the accuracy of the vehicle parking.

[0108] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "exemplary embodiment", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the exemplary description of the above terms does not necessarily mean the same embodiment or example.

[0109] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and purposes of the present application, and the scope of the present application is defined by the claims and their equivalents.

Claims

1. A method for identifying a corner point of a parking space, characterized by The method comprises the following steps: acquiring a parking space angle point coordinate of a parking vehicle in a panoramic image of the parking vehicle; determining a target area of the parking space according to the parking space angle point coordinate and a parking vehicle coordinate in the panoramic image; determining a near angle point coordinate in the parking space angle point coordinate, determining a region of interest of the parking space according to the near angle point coordinate and the parking vehicle coordinate, and taking the region of interest as the target area of the parking space; when a target parking space angle point is missing according to the target area, performing angle point restoration on a missing area of the target parking space angle point; performing edge extraction on the region of interest to obtain an edge line of the region of interest, determining a direction vector of the edge line according to the edge line, and determining the target parking space angle point according to the direction vector and a preset feature description vector; screening the direction vector according to a preset parking space mark line angle, determining a set of intersection points formed by the screened direction vector as a to-be-determined set of the parking space angle points, and performing feature matching on the to-be-determined set of the parking space angle points according to a preset feature description vector to obtain the target parking space angle point.

2. The method of identifying a corner of a parking space according to claim 1, characterized in that, The edge extraction on the region of interest comprises the following steps: performing edge extraction on the region of interest in an RGB channel.

3. The method of identifying a corner of a parking space according to claim 1, wherein The determination of the direction vector of the edge line comprises the following steps: performing a preset transformation on the edge line to obtain the direction vector of the edge line.

4. The method of identifying a corner of a parking space according to claim 1, wherein The feature matching on the to-be-determined set of the parking space angle points comprises the following steps: calculating a dot product of the preset feature description vector and a description vector in the to-be-determined set of the parking space angle points; taking an intersection point of the description vector in the to-be-determined set that satisfies a preset condition as the target parking space angle point.

5. The method of claim 1, wherein The determination of the target area of the parking space according to the parking space angle point coordinate and the parking vehicle coordinate in the panoramic image comprises the following steps: determining a parking space angle point coordinate in a preset confidence area, wherein the preset confidence area is a low confidence area; performing cropping on the panoramic image according to the parking space angle point coordinate in the preset confidence area and the parking vehicle coordinate to obtain the target area.

6. The method of identifying a corner of a parking space according to claim 5, wherein, The determination of the parking space angle point coordinate in the preset confidence area comprises the following steps: if the parking space angle point coordinate is in a set area range, determining that the parking space angle point coordinate is in the preset confidence area.

7. The method of identifying a corner of a parking space according to claim 5, wherein After the cropping on the panoramic image according to the parking space angle point coordinate in the preset confidence area and the parking vehicle coordinate to obtain the target area, the method further comprises the following steps: if the target area intersects with a vehicle area, performing area adjustment on the target area until the target area does not intersect with the vehicle area.

8. The method of identifying a corner of a parking space according to claim 7, wherein, The angle point restoration on the missing area of the target parking space angle point comprises the following steps: acquiring a parking space angle point coordinate adjacent to the target parking space angle point; determining geometric proportion data of the adjacent parking space angle point coordinate and the target parking space angle point; performing angle point restoration on the missing area of the target parking space angle point according to the geometric proportion data.

9. A storage medium, characterized by The storage medium has stored thereon a parking spot corner point identification program which, when executed by the processor, implements the parking spot corner point identification method according to any one of claims 1-8.

10. An electronic device, comprising: Comprising: A processor, a memory, and a parking spot corner point identification program stored on the memory and executable on the processor, which, when executed by the processor, implements the parking spot corner point identification method according to any one of claims 1-8.

11. A vehicle characterized by comprising: The electronic device of claim 10 is included.

Images