A parking information determination method, electronic device, storage medium, and program product

Abstract

This application provides a parking information determination method, electronic device, storage medium, and program product, aiming to improve the accuracy of parking information. The method includes: acquiring a first bird's-eye view and a second bird's-eye view, wherein the first bird's-eye view includes images taken from the front and rear views of the vehicle, and the second bird's-eye view includes images taken from the left and right views of the vehicle; stitching the first and second bird's-eye views together in terms of image channels to obtain a multi-channel image; inputting the multi-channel image into a parking information detection model, processing the multi-channel image through the parking information detection model, and obtaining parking information output by the parking information detection model, wherein the parking information is used to describe parking space-related information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and in particular to a parking information determination method, an electronic device, a storage medium and a program product. BACKGROUND

[0002] With the wide application of assisted parking and automatic parking technology, the accuracy requirement of parking information is increasingly improved.

[0003] In the related art, the parking information is mainly determined through a complete bird's eye view. However, in the splicing link of the complete bird's eye view, the overlapping parts between the bird's eye views under each view angle cannot be accurately aligned, resulting in that the bird's eye view obtained by splicing has information errors and losses, and further affecting the accuracy of the parking information. SUMMARY

[0004] In view of the above problems, the embodiments of the present application provide a parking information determination method, an electronic device, a storage medium and a program product, so as to overcome the above problems or at least partially solve the above problems.

[0005] In a first aspect, the embodiments of the present application provide a parking information determination method, comprising:

[0006] obtaining a first bird's eye view and a second bird's eye view, the first bird's eye view comprising image content of a front view angle and image content of a rear view angle of a vehicle, and the second bird's eye view comprising image content of a left view angle and image content of a right view angle of the vehicle;

[0007] splicing the first bird's eye view and the second bird's eye view on an image channel to obtain a multi-channel image;

[0008] inputting the multi-channel image into a parking information detection model, processing the multi-channel image through the parking information detection model, and obtaining parking information output by the parking information detection model, the parking information being used to describe parking space related information.

[0009] Optionally, the processing of the multi-channel image through the parking information detection model comprises:

[0010] detecting, through the parking information detection model, a suspected parking space position frame, a corner point position and a parking space occupancy attribute corresponding to each suspected parking space position frame in the multi-channel image, the parking space occupancy attribute being used to describe whether a parking space is occupied;

[0011] The parking information detection model is used to screen the suspected parking space position frame in the multi-channel image in combination with the corner point position, to obtain an actual parking space position frame in the multi-channel image, a corner point position corresponding to the actual parking space position frame, and a parking space occupancy attribute corresponding to the actual parking space position frame.

[0012] Optionally, the screening of the suspected parking space position frame in the multi-channel image in combination with the corner point position by the parking information detection model to obtain the actual parking space position frame in the multi-channel image comprises:

[0013] The left corner point and the right corner point corresponding to the suspected parking space position frame are predicted by the parking information detection model.

[0014] It is detected whether the corner point at the corner point position matches the left corner point and the right corner point.

[0015] The suspected parking space position frame that matches successfully is determined as the actual parking space position frame.

[0016] Optionally, the processing of the multi-channel image by the parking information detection model comprises:

[0017] It is detected by the parking information detection model whether an obstacle in the multi-channel image has a rotation angle relative to the multi-channel image.

[0018] For the obstacle having the rotation angle, a first obstacle position frame used to describe a ground position of the obstacle is determined by performing rotation target detection on the obstacle by the parking information detection model.

[0019] For the obstacle not having the rotation angle, a second obstacle position frame used to describe the ground position of the obstacle is determined by performing horizontal target detection on the obstacle by the parking information detection model.

[0020] Optionally, the determination of the first obstacle position frame used to describe the ground position of the obstacle by performing the rotation target detection on the obstacle by the parking information detection model comprises:

[0021] A center point of the obstacle and vectors from the center point to four edges of a first circumscribed rectangle of the obstacle are determined by the parking information detection model, and the first circumscribed rectangle has the rotation angle relative to the multi-channel image.

[0022] The first obstacle position frame is determined by the parking information detection model in combination with the center point and the vectors.

[0023] Optionally, the horizontal target detection of the obstacle by the parking information detection model determines a second obstacle position box used to describe a ground position of the obstacle, including:

[0024] The center point of the obstacle and the length and width of the second circumscribed rectangle of the obstacle are determined by the parking information detection model, and the second circumscribed rectangle does not have a rotation angle relative to the multi-channel image.

[0025] The second obstacle position box is determined by the parking information detection model in combination with the center point and the length and width of the second circumscribed rectangle.

[0026] Optionally, the processing of the multi-channel image by the parking information detection model includes:

[0027] The multi-channel image is semantically segmented by the parking information detection model to obtain a semantic segmentation result; and the semantic segmentation result includes at least one of type information of an object around the vehicle and an image region covered by the object in the multi-channel image.

[0028] Optionally, the first bird's eye view and the second bird's eye view are obtained, including:

[0029] An image size of a complete bird's eye view is obtained based on the image size of the complete bird's eye view.

[0030] The bird's eye picture of the front view and the bird's eye picture of the rear view are spliced in image positions based on the image size of the complete bird's eye view to obtain the first bird's eye view, and the bird's eye picture of the left view and the bird's eye picture of the right view are spliced in image positions based on the image size of the complete bird's eye view to obtain the second bird's eye view.

[0031] Optionally, the first bird's eye view and the second bird's eye view are both three-channel images, and the multi-channel image is a six-channel image.

[0032] The second aspect of the embodiment of the present application provides an electronic device, including a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to realize the parking information determination method disclosed in the embodiment of the present application.

[0033] The third aspect of the embodiment of the present application provides a computer readable storage medium, which stores a computer program / instruction, and the computer program / instruction is executed by a processor to realize the parking information determination method disclosed in the embodiment of the present application.

[0034] In a fourth aspect, the present application provides a computer program product, comprising computer programs / instructions, which, when executed by a processor, implement the parking information determination method disclosed in the present application.

[0035] The present application has the following advantages:

[0036] In the present application, by splicing the first bird's-eye view and the second bird's-eye view in the image channel, the registration and fusion of the overlapping area in the previous splicing process of the complete bird's-eye view (i.e. using the image content splicing method) can be avoided, so as to avoid the information error and loss, ensure that the obtained multi-channel image can accurately describe the image content of each view, and determine the parking information based on the information complete multi-channel image, thereby improving the accuracy of the parking information. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the description of the present application. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0038] Figure 1 is a step flow chart of a parking information determination method in the present application;

[0039] Figure 2 is a schematic diagram of an obstacle position box in the present application;

[0040] Figure 3 is a step flow chart of another parking information determination method in the present application;

[0041] Figure 4 is a schematic diagram of a first bird's-eye view in the present application;

[0042] Figure 5 is a schematic diagram of a second bird's-eye view in the present application;

[0043] Figure 6 is a structural schematic diagram of a parking information determination device in the present application;

[0044] Figure 7 is a schematic diagram of an electronic device in the present application. DETAILED DESCRIPTION

[0045] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0046] In recent years, important progress has been made in the research of computer vision, deep learning, machine learning, image processing, image recognition and other technologies based on artificial intelligence. Artificial intelligence (AI) is a new science and technology that studies and develops theories, methods, technologies and application systems for simulating and extending human intelligence. Artificial intelligence is a comprehensive discipline involving chips, big data, cloud computing, the Internet of Things, distributed storage, deep learning, machine learning, neural networks and many other technology categories. Computer vision, as an important branch of artificial intelligence, specifically enables machines to recognize the world. Computer vision technology generally includes face recognition, liveness detection, fingerprint recognition and anti-fraud verification, biometric recognition, face detection, pedestrian detection, object detection, pedestrian recognition, image processing, image recognition, image semantic understanding, image retrieval, character recognition, video processing, video content recognition, behavior recognition, three-dimensional reconstruction, virtual reality, augmented reality, simultaneous localization and mapping (SLAM), computational photography, robot navigation and positioning, and other technologies. With the research and progress of artificial intelligence technology, this technology has been applied in many fields, such as security, city management, traffic management, building management, park management, face passage, face attendance, logistics management, warehouse management, robots, intelligent marketing, computational photography, mobile imaging, cloud services, smart home, wearable devices, driverless vehicles, autonomous driving, intelligent medical care, face payment, face unlocking, fingerprint unlocking, face and certificate verification, smart screens, smart televisions, cameras, mobile Internet, network live streaming, beauty, makeup, medical cosmetology, intelligent temperature measurement, and other fields.

[0047] With the development of computer vision based on artificial intelligence, parking information determined based on images is widely used in assisted parking and automatic parking technologies, and plays an important role in identifying parking spaces near vehicles and planning parking routes. Currently, parking information is mainly determined through complete bird's eye view images. However, in the stitching process of complete bird's eye view images, the overlapping parts between bird's eye view images under different perspectives are difficult to accurately align, so that the bird's eye view images obtained after stitching have defects of information errors and loss compared with the bird's eye view images under different perspectives before stitching, thereby affecting the accuracy of the obtained parking information.

[0048] The embodiment of the present application provides a parking information determination method, as shown in Figure 1 The parking information determination method includes the following steps: Figure 1

[0049] ​Step S11: obtaining a first bird's eye view and a second bird's eye view, the first bird's eye view comprising image contents of a front view and a rear view of the vehicle, and the second bird's eye view comprising image contents of a left view and a right view of the vehicle.

[0050] Exemplarily, one vehicle-mounted camera is arranged at each of the front, rear, left and right body positions of the vehicle, and each vehicle-mounted camera is controlled to perform image capturing to obtain image contents of the front view, the rear view, the left view and the right view. The first bird's eye view is obtained based on the image contents of the front view and the rear view, and the second bird's eye view is obtained based on the image contents of the left view and the right view, so as to avoid the situation that the image contents overlap and need to be aligned in the process of obtaining a single bird's eye view corresponding to different views.

[0051] For example, for a bird's eye view around view containing image contents of all views, or a bird's eye view containing image contents of the front view and the left view, image registration needs to be performed on the overlapping parts of the image contents of different views, and then image fusion is performed on the overlapping parts to obtain a single bird's eye view corresponding to different views.

[0052] However, considering the image contents of the front view and the rear view, or the image contents of the left view and the right view, there is no overlapping of the image contents. Therefore, in the process of obtaining the first bird's eye view and the second bird's eye view, the steps of image registration and image fusion on the overlapping regions can be omitted, so that the first bird's eye view and the second bird's eye view recorded in the embodiments of the present application can correspond to complete image contents of the front view and the rear view, and complete image contents of the left view and the right view.

[0053] Step S12: splicing the first bird's eye view and the second bird's eye view in the image channel to obtain a multi-channel image.

[0054] As a possible implementation, single-channel images corresponding to the first bird's eye view and the second bird's eye view are obtained, each single-channel image is subjected to superposition processing or concatenation processing according to the channel type, and then each processed single-channel image is spliced according to the channel type to obtain a multi-channel image.

[0055] Optionally, the first bird's eye view and the second bird's eye view are each a three-channel image, and the multi-channel image is a six-channel image.

[0056] Exemplarily, assuming that the first bird's eye view and the second bird's eye view are both RGB three-channel images, for two pixel points of the same position in the first bird's eye view and the second bird's eye view, the two pixel points correspond to the six-channel data of the pixel point of the same position in the multi-channel image, that is, each pixel point in the multi-channel image has two RGB data, thereby obtaining a six-channel image.

[0057] Step S13: inputting the multi-channel image into a parking information detection model, processing the multi-channel image through the parking information detection model, and obtaining parking information output by the parking information detection model, the parking information being used to describe parking space related information.

[0058] It can be understood that the multi-channel image obtained through the channel splicing technology does not modify (such as deleting) the image content of the first bird's eye view and the second perspective bird's eye view, so that the multi-channel image can completely contain the image features in the first bird's eye view and the second perspective bird's eye view, thereby further avoiding the defects of information errors and losses existing in the complete bird's eye view.

[0059] The technical scheme of the embodiment of the present application splices the first bird's eye view and the second perspective bird's eye view on the image channel to obtain a multi-channel image used to determine the parking information, which can effectively represent the image content of the bird's eye view of each perspective, overcoming the defects of information errors and losses existing in the present technology that the overlapping parts between the bird's eye views of each perspective cannot be accurately aligned in the splicing link of the complete bird's eye view, thereby improving the accuracy of the parking information.

[0060] Optionally, the implementation process of processing the multi-channel image through the parking information detection model comprises:

[0061] detecting, through the parking information detection model, a suspected parking space position frame, a corner point position, and a parking space occupancy attribute corresponding to each suspected parking space position frame in the multi-channel image, the parking space occupancy attribute being used to describe whether the parking space is occupied;

[0062] screening, through the parking information detection model in combination with the corner point position, the suspected parking space position frame in the multi-channel image to obtain an actual parking space position frame in the multi-channel image, a corner point position corresponding to the actual parking space position frame, and a parking space occupancy attribute corresponding to the actual parking space position frame.

[0063] Exemplarily, the multi-channel image is input into the parking information detection model, the parking information detection model labels the parking space positions detected in the multi-channel image with a rectangular frame to obtain a plurality of initial parking space position frames and a plurality of initial corner point positions, and determines whether the initial parking space position frame corresponds to a parking space occupation attribute by detecting whether a non-parking space object exists in the labeled area, for example, if a non-parking space object exists in the labeled area, it is determined that the parking space is occupied, and if a non-parking space object does not exist in the labeled area, it is determined that the parking space is not occupied. The parking information detection model further screens and performs non-maximum suppression processing on the plurality of initial parking space position frames and initial corner point positions to obtain a suspected parking space position frame and a corner point position with higher detection accuracy.

[0064] The parking information detection model matches the suspected parking space position frame and the corner point position, for example, if a suspected parking space position frame can find a corner point position that meets the set condition from the corner point positions, it is determined that the suspected parking space position frame actually exists, and is determined as an actual parking space position frame, and if a suspected parking space position frame cannot find a corner point position that meets the set condition, it is determined that the suspected parking space position frame does not exist, and is determined as clear.

[0065] The set condition can be that the distance between at least two corner point positions and the suspected parking space position frame is less than a set threshold.

[0066] As a possible implementation, the left and right points at the entrance of the parking space are set as corner points (left corner point and right corner point).

[0067] In this embodiment, the parking information detection model is used to perform a parking space detection task, which detects parking spaces, corner points of the parking spaces, and whether the parking spaces are occupied.

[0068] Optionally, the implementation process of screening the suspected parking space position frame in the multi-channel image to obtain the actual parking space position frame in the multi-channel image by the parking information detection model in combination with the corner point position includes:

[0069] Predicting the left corner point and the right corner point corresponding to the suspected parking space position frame by the parking information detection model;

[0070] Detecting whether the corner point at the corner point position matches the left corner point and the right corner point;

[0071] Determining the suspected parking space position frame that matches successfully as the actual parking space position frame.

[0072] Exemplarily, after obtaining the suspected parking space position frame, the parking information detection model predicts the positions of the left and right corner points at the entrance of the parking space described by the suspected parking space position frame (i.e., predicts the positions of the two corner points at the entrance of the parking space), obtains the left and right corner points corresponding to the suspected parking space position frame, and then searches the detected corner point positions to determine whether there is a corner point position identical to the left and right corner points. If the corner point position exists, it is determined that the predicted left and right corner points actually exist, and it is further determined that the suspected parking space position frame actually exists. If the corner point position does not exist, it is determined that the suspected parking space position frame does not exist, thereby effectively improving the reliability of parking space detection.

[0073] Optionally, the implementation process of processing the multi-channel image by the parking information detection model comprises:

[0074] By the parking information detection model, it is detected whether the obstacle in the multi-channel image has a rotation angle relative to the multi-channel image.

[0075] For the obstacle having a rotation angle, the parking information detection model performs rotation target detection on the obstacle to determine a first obstacle position frame for describing the ground position of the obstacle.

[0076] For the obstacle not having a rotation angle, the parking information detection model performs horizontal target detection on the obstacle to determine a second obstacle position frame for describing the ground position of the obstacle.

[0077] In this embodiment, the parking information detection model is used to perform an obstacle target detection task. Compared with the traditional obstacle detection method of covering all regions (such as the column body region of a column and the contact region of the column and the ground) of the obstacle by semantic segmentation, and then determining the ground position (such as the position of the contact region of the column and the ground) of the obstacle from the segmented image region through post-processing, the obstacle target detection task performed by the embodiment divides the obstacle into an obstacle having a rotation angle (such as a column and a vehicle placed obliquely in the image) and an obstacle not having a rotation angle, and uses a rotation target detection method and a horizontal target detection method to directly mark the ground position of the obstacle by using an obstacle position frame (a first obstacle position frame and a second obstacle position frame), thereby avoiding the cumbersome step of determining the ground position of the obstacle through post-processing on the semantic segmentation result, and improving the detection accuracy of the ground position of the obstacle.

[0078] As one possible implementation, after inputting the multi-channel image into the parking information detection model, the parking information detection model will predict the confidence level of whether the obstacle has a rotation angle, and then perform different target detection processing according to the predicted confidence level. For example, if the obstacle is judged to have a rotation angle, rotation target detection will be performed, otherwise horizontal target detection will be performed.

[0079] Figure 2 A schematic diagram of the obstacle location box is shown. The origin of the coordinate axis in the diagram is the center point of the obstacle. I, II, III, and IV represent the coordinate quadrants. The solid-lined box represents the first bounding rectangle, and the dashed-lined box represents the second bounding rectangle. (t, r, l, b) are the vectors from the center point to the four sides of the first bounding rectangle, respectively. e ,h e Let the width and length of the second outer rectangle be defined respectively.

[0080] Optionally, the process of determining a first obstacle location box describing the ground position of the obstacle by performing rotating target detection on the obstacle using the parking information detection model includes:

[0081] The parking information detection model determines the center point of the obstacle and the vectors from the center point to the four sides of the first bounding rectangle of the obstacle, wherein the first bounding rectangle has a rotation angle relative to the multi-channel image.

[0082] The first obstacle location box is determined by combining the parking information detection model with the center point and the vector.

[0083] Understandably, assuming Figure 2 The coordinate axes shown are consistent with the image coordinate axes of the multi-channel image. It can be seen that the directions of the four sides of the first circumscribed rectangle are inconsistent with the direction of the image coordinate axes, and each side has an angle with the coordinate axis. This angle can be considered as the rotation angle of the first circumscribed rectangle and the obstacle corresponding to the first circumscribed rectangle.

[0084] Optionally, the process of performing horizontal target detection on the obstacle using the parking information detection model to determine a second obstacle location box for describing the ground position of the obstacle includes:

[0085] The parking information detection model is used to determine the center point of the obstacle, as well as the length and width of the second bounding rectangle of the obstacle. The second bounding rectangle has no rotation angle relative to the multi-channel image.

[0086] The second obstacle location box is determined by using the parking information detection model, combined with the center point and the length and width of the second circumscribed rectangle.

[0087] Optionally, the process of processing the multi-channel image using the parking information detection model includes:

[0088] The parking information detection model is used to perform semantic segmentation on the multi-channel image to obtain a semantic segmentation result; wherein, the semantic segmentation result includes the type information of objects around the vehicle, and at least one of the image regions covered by the objects in the multi-channel image.

[0089] In this embodiment, the parking information detection model is used to perform a semantic segmentation task, which is used to perform semantic segmentation on various objects in multi-channel images, such as lane lines and vehicles, to obtain image regions that describe the positions of the objects.

[0090] For example, such as Figure 3 The flowchart of another parking information determination method is shown. The first and second bird's-eye view images are first stitched together into a multi-channel image; then the multi-channel image is processed by a feature extraction network to extract features, resulting in a multi-scale pyramid feature set (i.e., multi-scale features); these multi-scale features are then fused by a feature fusion network to obtain the fused target feature image.

[0091] The aforementioned target feature images are input into a task detection model, which performs at least one of the following tasks: parking space detection, semantic segmentation, and obstacle detection. Each task takes the target feature image as input data and outputs a result. The results output by the task detection model constitute the parking information.

[0092] The two models used for feature extraction and feature fusion can be implemented independently of the task detection model using models from related technologies, or they can be integrated into the task detection model and trained to perform tasks such as parking space detection, semantic segmentation, and obstacle detection. This application does not limit the type and number of models in the task detection model.

[0093] As one possible implementation, the above-mentioned task detection model may include the parking information detection model provided in the embodiments of this application. The parking information detection model can perform at least one of the following tasks: obstacle target detection task (i.e. obstacle detection task), parking space detection task, and semantic segmentation task. The remaining tasks in the multi-task detection model can be performed by a model with corresponding task detection function in related technologies.

[0094] Optionally, the process of obtaining the first bird's-eye view and the second bird's-eye view includes:

[0095] acquire the bird's eye view image of the front view angle of the vehicle, the bird's eye view image of the rear view angle of the vehicle, the bird's eye view image of the left view angle of the vehicle and the bird's eye view image of the right view angle of the vehicle, and acquire the image size of the complete bird's eye view around view;

[0096] based on the image size of the complete bird's eye view around view, the bird's eye view picture of the front view angle and the bird's eye view picture of the rear view angle are spliced in the image position to obtain the first bird's eye view, and based on the image size of the complete bird's eye view around view, the bird's eye view picture of the left view angle and the bird's eye view picture of the right view angle are spliced in the image position to obtain the second bird's eye view.

[0097] Figure 4 and Figure 5 The schematic diagrams of the first bird's eye view and the second bird's eye view are shown respectively. As can be seen from the schematic diagrams, when the bird's eye views of the front view angle, the rear view angle, the left view angle and the right view angle are spliced respectively, the relative positions of the bird's eye views of the respective view angles are preserved, that is, the bird's eye view of the front view angle is placed above the first bird's eye view when spliced, the bird's eye view of the rear view angle is placed below the first bird's eye view when spliced, the bird's eye view of the left view angle is placed to the left of the second bird's eye view when spliced, and the bird's eye view of the right view angle is placed to the right of the second bird's eye view when spliced, so that the image sizes of the first bird's eye view and the second bird's eye view are consistent with the image size of the complete bird's eye view around view spliced from the bird's eye views of the same view angles, so as to facilitate subsequent model input. Moreover, the first bird's eye view and the second bird's eye view preserve the view angle position relationship of the image contents of the respective view angles, and can further avoid the generation of overlapping areas.

[0098] As a possible implementation manner, the bird's eye view images of the respective view angles are acquired by the following manner:

[0099] An fisheye camera is arranged on the vehicle, and fisheye camera images of the respective view angles are obtained by shooting. A distortion correction matrix is calculated according to the intrinsic matrix and the distortion coefficient of the fisheye camera, so as to project the information in the fisheye camera images into non-distortion images.

[0100] According to the four-point coordinates respectively determined in the non-distortion images of the respective view angles and the corresponding four-point coordinates in the bird's eye views of the respective view angles, a conversion matrix is determined, so as to project the non-distortion images of the respective view angles into the bird's eye views of the respective view angles.

[0101] It should be noted that, for the method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the embodiments of the present application are not limited by the action order described, because according to the embodiments of the present application, certain steps can be performed in other order or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions involved are not necessarily the necessary of the embodiments of the present application.

[0102] Figure 6 Fig. 1 is a structural schematic diagram of a parking information determination device according to an embodiment of the present application. The parking information determination device comprises an acquisition module, a processing module and a detection module, wherein:

[0103] The acquisition module is configured to acquire a first bird's eye view and a second bird's eye view. The first bird's eye view comprises image content of a front view and image content of a rear view of a vehicle. The second bird's eye view comprises image content of a left view and image content of a right view of the vehicle.

[0104] The processing module is configured to splice the first bird's eye view and the second bird's eye view to obtain a multi-channel image.

[0105] The detection module is configured to input the multi-channel image into a parking information detection model, process the multi-channel image by using the parking information detection model, and obtain parking information output by the parking information detection model. The parking information is used to describe parking space related information.

[0106] Optionally, the detection module is specifically configured to detect, by using the parking information detection model, a suspected parking space position frame, a corner point position and a parking space occupancy attribute corresponding to each suspected parking space position frame in the multi-channel image. The parking space occupancy attribute is used to describe whether a parking space is occupied.

[0107] The detection module is configured to screen the suspected parking space position frame in the multi-channel image by using the parking information detection model in combination with the corner point position, to obtain an actual parking space position frame in the multi-channel image, a corner point position corresponding to the actual parking space position frame and a parking space occupancy attribute corresponding to the actual parking space position frame.

[0108] Optionally, the detection module is further configured to predict, by using the parking information detection model, a left corner point and a right corner point corresponding to the suspected parking space position frame, detect whether the corner point at the corner point position matches the left corner point and the right corner point, and determine a suspected parking space position frame that matches successfully as the actual parking space position frame.

[0109] Optionally, the detection module is further configured to detect, by using the parking information detection model, whether an obstacle in the multi-channel image has a rotation angle relative to the multi-channel image.

[0110] For the obstacle having the rotation angle, the detection module is configured to perform rotation target detection on the obstacle by using the parking information detection model, to determine a first obstacle position frame used to describe a ground position of the obstacle.

[0111] For the obstacle without the rotation angle, the parking information detection model is used for horizontal target detection on the obstacle to determine a second obstacle position frame used for describing a ground position of the obstacle.

[0112] Optionally, the detection module is further configured to determine, by the parking information detection model, a center point of the obstacle and vectors from the center point to four edges of a first circumscribed rectangle of the obstacle, the first circumscribed rectangle having a rotation angle relative to the multi-channel image.

[0113] The first obstacle position frame is determined by the parking information detection model in combination with the center point and the vectors.

[0114] Optionally, the detection module is further configured to determine, by the parking information detection model, a center point of the obstacle and a length and a width of a second circumscribed rectangle of the obstacle, the second circumscribed rectangle having no rotation angle relative to the multi-channel image.

[0115] The second obstacle position frame is determined by the parking information detection model in combination with the center point and the length and the width of the second circumscribed rectangle.

[0116] Optionally, the detection module is further configured to perform semantic segmentation on the multi-channel image by the parking information detection model to obtain a semantic segmentation result, wherein the semantic segmentation result includes at least one of type information of an object around the vehicle and an image region covered by the object in the multi-channel image.

[0117] Optionally, the acquisition module is specifically configured to acquire a bird's-eye view image of a front view angle of the vehicle, a bird's-eye view image of a rear view angle of the vehicle, a bird's-eye view image of a left view angle of the vehicle, and a bird's-eye view image of a right view angle of the vehicle, and acquire an image size of a complete bird's-eye view.

[0118] The bird's-eye view image of the front view angle and the bird's-eye view image of the rear view angle are spliced in image positions based on the image size of the complete bird's-eye view to obtain the first bird's-eye view, and the bird's-eye view image of the left view angle and the bird's-eye view image of the right view angle are spliced in image positions based on the image size of the complete bird's-eye view to obtain the second bird's-eye view.

[0119] Optionally, the first bird's-eye view and the second bird's-eye view are both three-channel images, and the multi-channel image is a six-channel image.

[0120] It should be noted that the device embodiment is similar to the method embodiment, and thus the description is relatively simple, and the related parts can be referred to the method embodiment.

[0121] The embodiment of the present application further provides an electronic device, referring to Figure 7 , Figure 7 is a schematic diagram of the electronic device provided by the embodiment of the present application. As shown in Figure 7 , the electronic device 100 comprises a memory 110 and a processor 120, the memory 110 and the processor 120 are in communication connection through a bus, and the memory 110 stores a computer program, the computer program can run on the processor 120, and then the steps in the parking information determination method disclosed by the embodiment of the present application are realized.

[0122] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program / instruction, and the computer program / instruction is executed by a processor to realize the parking information determination method disclosed by the embodiment of the present application.

[0123] The embodiment of the present application further provides a computer program product, which comprises a computer program / instruction, and the computer program / instruction is executed by a processor to realize the parking information determination method disclosed by the embodiment of the present application.

[0124] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same and similar parts between each embodiment can be referred to each other.

[0125] Those skilled in the art should understand that the embodiments of the embodiment of the present application can be provided as a method, device or computer program product. Therefore, the embodiment of the present application can adopt a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiment of the present application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0126] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the method, device, electronic device and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing terminal equipment to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal equipment realize the functions specified in the flowcharts and / or block diagrams. Figure 1 The device for realizing the functions specified in one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0127] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more flow or block Figure 1 one or more flow or block

[0128] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flow or block Figure 1 one or more flow or block

[0129] Although preferred embodiments of the application have been described, those skilled in the art will recognize that additional modifications and changes can be made thereto without departing from the scope of the present application. Accordingly, the appended claims are intended to cover all such modifications and changes as fall within the scope of the application.

[0130] Finally, it should be noted that the terms "first", "second", and the like, herein do not denote any order, quantity, combination, or importance, but rather are used to distinguish one element from another, and do not imply singular or plural. Also, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0131] The above provides a parking information determination method, electronic device, storage medium and program product, the principle and implementation mode of the present application are described by applying specific examples in the text, the above example is only used to help understand the method and core idea of the present application; At the same time, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed, and the above description should not be understood as the limitation of the present application.

Claims

1. A parking information determination method characterized by comprising: The method comprises: obtaining a first bird's-eye view and a second bird's-eye view, the first bird's-eye view comprising photograph image content of a front view angle and photograph image content of a rear view angle of a vehicle, and the second bird's-eye view comprising photograph image content of a left view angle and photograph image content of a right view angle of the vehicle; stitching the first bird's-eye view and the second bird's-eye view on an image channel to obtain a multi-channel image; inputting the multi-channel image into a parking information detection model, processing the multi-channel image through the parking information detection model, and obtaining parking information output by the parking information detection model, the parking information being used to describe parking space related information; wherein the obtaining of the first bird's-eye view and the second bird's-eye view comprises: obtaining a bird's-eye view image of the front view angle of the vehicle, a bird's-eye view image of the rear view angle of the vehicle, a bird's-eye view image of the left view angle of the vehicle, and a bird's-eye view image of the right view angle of the vehicle, and obtaining an image size of a complete bird's-eye view image; stitching the bird's-eye view image of the front view angle and the bird's-eye view image of the rear view angle on an image position based on the image size of the complete bird's-eye view image to obtain the first bird's-eye view, and stitching the bird's-eye view image of the left view angle and the bird's-eye view image of the right view angle on an image position based on the image size of the complete bird's-eye view image to obtain the second bird's-eye view; wherein the first bird's-eye view and the second bird's-eye view are both three-channel images, and the multi-channel image is a six-channel image.

2. The method of claim 1, wherein, The processing of the multi-channel image through the parking information detection model comprises: detecting, through the parking information detection model, a suspected parking space position box, a corner point position, and a parking space occupancy attribute corresponding to each suspected parking space position box in the multi-channel image, the parking space occupancy attribute being used to describe whether a parking space is occupied; screening, through the parking information detection model in combination with the corner point position, the suspected parking space position box in the multi-channel image to obtain an actual parking space position box in the multi-channel image, a corner point position corresponding to the actual parking space position box, and a parking space occupancy attribute corresponding to the actual parking space position box.

3. The method of claim 2, wherein, The screening, through the parking information detection model in combination with the corner point position, of the suspected parking space position box in the multi-channel image to obtain the actual parking space position box in the multi-channel image comprises: predicting, through the parking information detection model, a left corner point and a right corner point corresponding to the suspected parking space position box; detecting whether the corner point at the corner point position matches the left corner point and the right corner point; determining a suspected parking space position box that matches successfully as the actual parking space position box.

4. The method of claim 1, wherein, The processing of the multi-channel image through the parking information detection model comprises: detecting, through the parking information detection model, whether an obstacle in the multi-channel image has a rotation angle relative to the multi-channel image; for the obstacle having the rotation angle, performing rotation target detection on the obstacle through the parking information detection model to determine a first obstacle position box used to describe a ground position of the obstacle; For the obstacle without the rotation angle, the parking information detection model is used to perform horizontal target detection on the obstacle to determine a second obstacle position box used to describe a ground position of the obstacle.

5. The method of claim 4, wherein, The rotation target detection on the obstacle by the parking information detection model to determine the first obstacle position box used to describe the ground position of the obstacle includes: The parking information detection model is used to determine a center point of the obstacle and vectors from the center point to four edges of a first circumscribed rectangle of the obstacle, the first circumscribed rectangle having a rotation angle relative to the multi-channel image; The parking information detection model is used to determine the first obstacle position box in combination with the center point and the vectors.

6. The method of claim 4, wherein The horizontal target detection on the obstacle by the parking information detection model to determine the second obstacle position box used to describe the ground position of the obstacle includes: The parking information detection model is used to determine a center point of the obstacle and a length and a width of a second circumscribed rectangle of the obstacle, the second circumscribed rectangle not having a rotation angle relative to the multi-channel image; The parking information detection model is used to determine the second obstacle position box in combination with the center point and the length and the width of the second circumscribed rectangle.

7. The method according to any of claims 1 to 6, characterized in that The processing of the multi-channel image by the parking information detection model includes: The parking information detection model is used to perform semantic segmentation on the multi-channel image to obtain a semantic segmentation result, wherein the semantic segmentation result includes at least one of type information of an object around the vehicle and an image region covered by the object in the multi-channel image.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program comprises instructions that, when executed by the processor, cause the electronic device to perform the method of any one of claims 1-7. The processor executes the computer program to implement the parking information determination method in any one of claims 1 to 7.

9. A computer readable storage medium having stored thereon computer programs / instructions, characterized in that, The computer program / instruction is executed by the processor to implement the parking information determination method in any one of claims 1 to 7.

10. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instruction is executed by the processor to implement the parking information determination method in any one of claims 1 to 7. The computer program / instruction is executed by the processor to implement the parking information determination method in any one of claims 1 to 7.

Images