Parking space marking method, electronic device, vehicle and computer-readable storage medium
By using a coordinate conversion matrix in parking space labeling to fuse the annotation information under different coordinate systems, the problem of low accuracy in parking space labeling in the existing technology is solved, and the effect of automatic parking technology is improved.
Patent Information
- Application Number
- CN202411050190.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-01
AI Technical Summary
In the prior art, parking space marking is not accurate, which affects the effectiveness of automatic parking technology.
通过获取多个泊车位图像的标注信息,利用坐标转换矩阵将不同坐标系下的标注信息转换为统一坐标系,融合标注信息以提高标注准确度。
It improves the accuracy of parking space marking and enhances the reliability and efficiency of automatic parking technology.
Smart Images

Figure CN118570771B_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the field of intelligent vehicle technology, and specifically relates to a parking space marking method, electronic equipment, vehicle and computer-readable storage medium. Background Art
[0002] With the continuous development of vehicle intelligence, automatic parking technology, as a key function in vehicle intelligence, has received more and more attention.
[0003] Automatic parking technology generally detects parking space bird's-eye view images through parking space detection models to determine available parking spaces. However, in order to train the parking space detection model, a large amount of comprehensive parking space bird's-eye view data with parking space annotations is required. In the prior art, each parking space bird's-eye view image is annotated by using a pre-annotated model.
[0004] However, the marking accuracy of parking spaces using the existing methods is not high. Summary of the invention
[0005] The embodiments of the present application provide a parking space marking method, electronic device, vehicle and computer-readable storage medium, which can solve the problem of low accuracy of parking space marking in the prior art.
[0006] In a first aspect, an embodiment of the present application provides a parking space marking method, comprising:
[0007] Acquire first annotation information of a first parking space image, wherein the first parking space image is a parking space image to be annotated, and the first annotation information is annotated based on a first coordinate system;
[0008] acquiring third annotation information of the first parking space image based on second annotation information of the second parking space image and a first coordinate conversion matrix corresponding to the first parking space image, wherein the second parking space image and the first parking space image are collected in the same parking lot, the second annotation information is annotated based on a second coordinate system, the first coordinate conversion matrix is used to convert the annotation information of the second coordinate system to the first coordinate system, and the third annotation information is annotation information based on the first coordinate system;
[0009] First target annotation information of the first parking space image is obtained based on the first annotation information of the first parking space image and the third annotation information.
[0010] Optionally, the acquiring third annotation information of the first parking space image based on the second annotation information of the second parking space image and a first coordinate conversion matrix corresponding to the first parking space image includes:
[0011] acquiring second annotation information of the first parking space image based on second annotation information of the second parking space image;
[0012] Based on a first coordinate conversion matrix corresponding to the first parking space image and second annotation information of the first parking space image, third annotation information of the first parking space image is acquired.
[0013] Optionally, before acquiring the third annotation information of the first parking space image based on the first coordinate conversion matrix corresponding to the first parking space image and the second annotation information of the first parking space image, the method further includes:
[0014] Acquire first coordinate information of a collection device in the first parking space image, wherein the first coordinate information is coordinate information based on the first coordinate system;
[0015] Acquire second coordinate information of the acquisition device in the first parking space image, wherein the second coordinate information is coordinate information based on the second coordinate system;
[0016] A first coordinate conversion matrix corresponding to the first parking space image is acquired based on the first coordinate information and the second coordinate information.
[0017] Optionally, the acquiring second coordinate information of the acquisition device in the first parking space image includes:
[0018] Based on the correspondence between the first parking space image and the position information of the acquisition device in the second coordinate system, second coordinate information of the acquisition device in the first parking space image is acquired, wherein the position information of the acquisition device in the second coordinate system is the coordinate information of the acquisition device in the second coordinate system when the acquisition device acquires the first parking space image.
[0019] Optionally, the acquiring second coordinate information of the acquisition device in the first parking space image based on the correspondence between the first parking space image and the position information of the acquisition device in the second coordinate system includes:
[0020] Based on the identification information of the first parking space image, acquiring position information of the acquisition device corresponding to the identification information in the second coordinate system;
[0021] Based on the position information of the acquisition device in the second coordinate system corresponding to the identification information, second coordinate information of the acquisition device in the first parking space image is determined.
[0022] Optionally, the acquiring, based on the identification information of the first parking space image, the position information of the acquisition device corresponding to the identification information in the second coordinate system includes:
[0023] Based on the identification information, acquiring position information of the acquisition device corresponding to the identification information in a third coordinate system, and third coordinate information of a coordinate origin of the third coordinate system in the second coordinate system;
[0024] Based on the position information of the acquisition device in the third coordinate system and the third coordinate information of the coordinate origin of the third coordinate system in the second coordinate system, the position information of the acquisition device in the second coordinate system is acquired.
[0025] Optionally, the acquiring the second annotation information of the first parking space image based on the second annotation information of the second parking space image includes:
[0026] Acquire a mapping point in the first parking space image of a coordinate value corresponding to the second annotation information of the second parking space image;
[0027] Second annotation information of the first parking space image is determined based on the coordinate values of the mapping points in the first parking space image.
[0028] Optionally, the step of acquiring a coordinate value corresponding to the second annotation information of the second parking space image that is before a mapping point in the first parking space image further includes:
[0029] Acquire first annotation information of the second parking space image;
[0030] Based on the first annotation information of the second parking space image and a second coordinate conversion matrix corresponding to the second parking space image, second annotation information of the second parking space image is acquired, wherein the second coordinate conversion matrix is used to convert the annotation information of the first coordinate system into the second coordinate system.
[0031] Optionally, before acquiring the second annotation information of the second parking space image based on the first annotation information of the second parking space image and the second coordinate transformation matrix corresponding to the second parking space image, the method further includes:
[0032] Acquire fourth coordinate information of a collection device in the second parking space image, where the fourth coordinate information is coordinate information based on the first coordinate system;
[0033] Acquire fifth coordinate information of the acquisition device in the second parking space image, wherein the fifth coordinate information is coordinate information based on the second coordinate system;
[0034] A second coordinate conversion matrix corresponding to the second parking space image is obtained based on the fourth coordinate information and the fifth coordinate information.
[0035] Optionally, obtaining first target annotation information of the first parking space image based on the first annotation information of the first parking space image and the third annotation information includes:
[0036] According to a preset rule, determining that the third annotation information of the first parking space image is within the first parking space image;
[0037] For each corner point of the parking space, if both the first annotation information and the third annotation information exist, the first annotation information and the third annotation information are fused to obtain the first target annotation information of the corner point; if only the first annotation information exists, the first annotation information is determined to be the first target annotation information of the corner point; if only the third annotation information exists, the third annotation information is determined to be the first target annotation information of the corner point.
[0038] Optionally, also include:
[0039] According to the preset rule, if it is determined that the third annotation information of the first parking space image is not within the first parking space image, then the first annotation information of the first parking space image is determined to be the first target annotation information of the first parking space image.
[0040] Optionally, obtaining first target annotation information of the first parking space image based on the first annotation information of the first parking space image and the third annotation information includes:
[0041] If the first parking space image does not have the first annotation information, the third annotation information is determined to be the first target annotation information of the first parking space image.
[0042] Optionally, the first annotation information is obtained based on image recognition.
[0043] Optionally, the acquiring first annotation information of the first parking space image includes:
[0044] The first parking space image is input into a pre-labeling model to obtain first labeling information of the first parking space image.
[0045] Optionally, the first coordinate system is a pixel coordinate system, and the second coordinate system is an earth coordinate system.
[0046] Optionally, the identification information is timestamp information.
[0047] Optionally, the determining the second annotation information of the first parking space image based on the coordinate value of the mapping point in the first parking space image includes:
[0048] Dividing the mapping points in the first parking space image into ranges, and fusing the coordinate values of at least one mapping point in the same range to obtain a plurality of fused points;
[0049] Second annotation information of the first parking space image is obtained based on the multiple fusion points.
[0050] Optionally, also include:
[0051] Based on the first target annotation information of the first parking space image, acquiring fourth annotation information of the first parking space image, wherein the fourth annotation information is annotated based on the second coordinate system;
[0052] acquiring fifth annotation information of the third parking space image based on a mapping point of fourth annotation information of the first parking space image on the third parking space image, wherein the fifth annotation information is annotated based on the second coordinate system, wherein the third parking space image and the first parking space image are parking spaces of the same parking lot collected under different collection conditions;
[0053] Based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image, second target annotation information of the third parking space image is obtained, wherein the second target annotation information is annotated based on the first coordinate system.
[0054] Optionally, also include:
[0055] Acquire sixth annotation information of the third parking space image, wherein the sixth annotation information is annotated based on the first coordinate system;
[0056] The obtaining the second target annotation information of the third parking space image based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image includes:
[0057] Based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image, obtaining seventh annotation information of the third parking space image, wherein the seventh annotation information is annotated based on the first coordinate system;
[0058] Based on the sixth labeling information and the seventh labeling information, second target labeling information of the third parking space image is obtained.
[0059] Optionally, obtaining second target annotation information of the third parking space image based on the sixth annotation information and the seventh annotation information includes:
[0060] For the first type of corner points: the sixth annotation information and the seventh annotation information are merged to obtain second target annotation information of the first type of corner points, wherein the first type of corner points are corner points at both ends of the parking space entrance line;
[0061] For the second type of corner points: determine the seventh annotation information as the second target annotation information of the second type of corner points, wherein the second type of corner points are the corner points at both ends of the bottom edge line of the parking space.
[0062] Optionally, also include:
[0063] Acquire sample data, the sample data comprising: a parking space image sample, and position information of the acquisition device in a second coordinate system when acquiring the parking space image sample, wherein the parking space image sample comprises: the first parking space image and the second parking space image.
[0064] Optionally, the position information of the acquisition device in the second coordinate system includes: coordinate information of the acquisition device in the second coordinate system.
[0065] Optionally, the position information of the acquisition device in the second coordinate system includes: coordinate information of the acquisition device in a third coordinate system, and coordinate information of the origin of the third coordinate system in the second coordinate system.
[0066] Optionally, the parking space image samples and the position information of the acquisition device in the second coordinate system when the parking space image samples are acquired have a corresponding relationship.
[0067] Optionally, the parking space image samples are named with identification information, and the position information of the acquisition device in the second coordinate system when acquiring the parking space image samples corresponds to the identification information.
[0068] In a second aspect, an embodiment of the present application provides a parking space marking device, including:
[0069] an acquisition module, configured to acquire first annotation information of a first parking space image, wherein the first parking space image is a parking space image to be annotated, and the first annotation information is annotated based on a first coordinate system;
[0070] a processing module, configured to obtain third annotation information of the first parking space image based on second annotation information of the second parking space image and a first coordinate conversion matrix corresponding to the first parking space image, wherein the second parking space image and the first parking space image are collected in the same parking lot, the second annotation information is annotated based on a second coordinate system, the first coordinate conversion matrix is used to convert the annotation information of the second coordinate system into the first coordinate system, and the third annotation information is annotation information based on the first coordinate system; and obtain first target annotation information of the first parking space image based on the first annotation information and the third annotation information.
[0071] Optionally, the processing module is specifically used to obtain second annotation information of the first parking space image based on second annotation information of the second parking space image; and obtain third annotation information of the first parking space image based on a first coordinate transformation matrix corresponding to the first parking space image and the second annotation information of the first parking space image.
[0072] Optionally, the processing module is further used to obtain first coordinate information of the acquisition device in the first parking space image, wherein the first coordinate information is coordinate information based on the first coordinate system; obtain second coordinate information of the acquisition device in the first parking space image, wherein the second coordinate information is coordinate information based on the second coordinate system; and obtain a first coordinate conversion matrix corresponding to the first parking space image based on the first coordinate information and the second coordinate information.
[0073] Optionally, the processing module is specifically used to obtain second coordinate information of the acquisition device in the first parking space image based on the correspondence between the first parking space image and the position information of the acquisition device in the second coordinate system, wherein the position information of the acquisition device in the second coordinate system is the coordinate information of the acquisition device in the second coordinate system when the acquisition device acquires the first parking space image.
[0074] Optionally, the processing module is specifically configured to obtain, based on identification information of the first parking space image, position information of the acquisition device corresponding to the identification information in the second coordinate system;
[0075] Based on the position information of the acquisition device in the second coordinate system corresponding to the identification information, second coordinate information of the acquisition device in the first parking space image is determined.
[0076] The optional processing module is specifically used to obtain, based on the identification information, the position information of the acquisition device corresponding to the identification information in the third coordinate system, and the third coordinate information of the coordinate origin of the third coordinate system in the second coordinate system;
[0077] Based on the position information of the acquisition device in the third coordinate system and the third coordinate information of the coordinate origin of the third coordinate system in the second coordinate system, the position information of the acquisition device in the second coordinate system is acquired.
[0078] Optionally, the processing module is specifically configured to obtain a mapping point in the first parking space image of a coordinate value corresponding to the second annotation information of the second parking space image;
[0079] Second annotation information of the first parking space image is determined based on the coordinate values of the mapping points in the first parking space image.
[0080] Optionally, the processing module is further used to obtain first annotation information of the second parking space image;
[0081] Based on the first annotation information of the second parking space image and a second coordinate conversion matrix corresponding to the second parking space image, second annotation information of the second parking space image is acquired, wherein the second coordinate conversion matrix is used to convert the annotation information of the first coordinate system into the second coordinate system.
[0082] Optionally, the processing module is further used to obtain fourth coordinate information of a collection device in the second parking space image, where the fourth coordinate information is coordinate information based on the first coordinate system;
[0083] Acquire fifth coordinate information of the acquisition device in the second parking space image, wherein the fifth coordinate information is coordinate information based on the second coordinate system;
[0084] A second coordinate conversion matrix corresponding to the second parking space image is obtained based on the fourth coordinate information and the fifth coordinate information.
[0085] Optionally, the processing module is specifically configured to determine, according to a preset rule, that the third annotation information of the first parking space image is within the first parking space image;
[0086] For each corner point of the parking space, if both the first annotation information and the third annotation information exist, the first annotation information and the third annotation information are fused to obtain the first target annotation information of the corner point; if only the first annotation information exists, the first annotation information is determined to be the first target annotation information of the corner point; if only the third annotation information exists, the third annotation information is determined to be the first target annotation information of the corner point.
[0087] Optionally, the processing module is further configured to determine, according to the preset rule, that if the third annotation information of the first parking space image is not within the first parking space image, then determine that the first annotation information of the first parking space image is the first target annotation information of the first parking space image.
[0088] Optionally, the processing module is specifically configured to determine that the third annotation information is first target annotation information of the first parking space image if the first annotation information does not exist in the first parking space image.
[0089] Optionally, the first annotation information is obtained based on image recognition.
[0090] Optionally, the acquisition module is specifically configured to input the first parking space image into a pre-labeling model to obtain first labeling information of the first parking space image.
[0091] Optionally, the first coordinate system is a pixel coordinate system, and the second coordinate system is an earth coordinate system.
[0092] Optionally, the identification information is timestamp information.
[0093] Optionally, the processing module is specifically used to divide the mapping points in the first parking space image into ranges, and perform fusion processing on the coordinate values of at least one mapping point in the same range to obtain a plurality of fusion points;
[0094] Second annotation information of the first parking space image is obtained based on the multiple fusion points.
[0095] Optionally, the processing module is further configured to obtain fourth annotation information of the first parking space image based on the first target annotation information of the first parking space image, wherein the fourth annotation information is annotated based on the second coordinate system;
[0096] acquiring fifth annotation information of the third parking space image based on a mapping point of fourth annotation information of the first parking space image on the third parking space image, wherein the fifth annotation information is annotated based on the second coordinate system, wherein the third parking space image and the first parking space image are parking spaces of the same parking lot collected under different collection conditions;
[0097] Based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image, second target annotation information of the third parking space image is obtained, wherein the second target annotation information is annotated based on the first coordinate system.
[0098] Optionally, the acquisition module is further used to acquire sixth annotation information of the third parking space image, wherein the sixth annotation information is annotated based on the first coordinate system;
[0099] The processing module is specifically configured to obtain seventh annotation information of the third parking space image based on fifth annotation information of the third parking space image and a first coordinate conversion matrix corresponding to the third parking space image, wherein the seventh annotation information is annotated based on the first coordinate system;
[0100] Based on the sixth labeling information and the seventh labeling information, second target labeling information of the third parking space image is obtained.
[0101] Optionally, the processing module is specifically configured to: for a first type of corner point: fuse the sixth annotation information and the seventh annotation information to obtain second target annotation information of the first type of corner point, wherein the first type of corner point is a corner point at both ends of a parking space entrance line;
[0102] For the second type of corner points: determine the seventh annotation information as the second target annotation information of the second type of corner points, wherein the second type of corner points are the corner points at both ends of the bottom edge line of the parking space.
[0103] Optionally, it also includes: an acquisition module, used to acquire sample data, the sample data including: parking space image samples, and position information of the acquisition device in the second coordinate system when acquiring the parking space image samples, wherein the parking space image samples include: the first parking space image and the second parking space image.
[0104] Optionally, the position information of the acquisition device in the second coordinate system includes: coordinate information of the acquisition device in the second coordinate system.
[0105] Optionally, the position information of the acquisition device in the second coordinate system includes: coordinate information of the acquisition device in a third coordinate system, and coordinate information of the origin of the third coordinate system in the second coordinate system.
[0106] Optionally, the parking space image samples and the position information of the acquisition device in the second coordinate system when the parking space image samples are acquired have a corresponding relationship.
[0107] Optionally, the parking space image samples are named with identification information, and the position information of the acquisition device in the second coordinate system when acquiring the parking space image samples corresponds to the identification information.
[0108] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the parking space marking method as described in any one of the first aspects are implemented.
[0109] In a fourth aspect, an embodiment of the present application provides a vehicle, comprising: a processor and a memory, wherein the memory stores a program or instruction that can be executed on the processor, and when the program or instruction is executed by the processor, the steps of the parking space marking method as described in any one of the first aspects are implemented.
[0110] In a fifth aspect, an embodiment of the present application provides a cloud server, comprising: a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the parking space marking method as described in any one of the first aspects are implemented.
[0111] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the parking space marking method as described in any one of the first aspects are implemented.
[0112] In a seventh aspect, an embodiment of the present application provides a computer program product, which, when executed by a processor of a vehicle or a cloud server, implements the steps of the parking space marking method as described in any one of the first aspects.
[0113] The parking space labeling method, electronic device, vehicle and computer-readable storage medium provided in the embodiments of the present application obtain first labeling information of a first parking space image, obtain third labeling information of the first parking space image based on the second labeling information of the second parking space image and a first coordinate conversion matrix corresponding to the first parking space image, and obtain first target labeling information of the first parking space image based on the first labeling information and the third labeling information of the first parking space image, wherein the first labeling information and the third labeling information are labeling information obtained in two different ways, and therefore the first target labeling information obtained by combining the first labeling information and the third labeling information has higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0114] Figure 1 A schematic diagram of the structure of a parking space sample data collection device provided in an embodiment of the present application;
[0115] Figure 2 A schematic diagram of a bird's-eye view of a parking space provided in an embodiment of the present application;
[0116] Figure 3A schematic diagram of a parking space marking method provided in an embodiment of the present application;
[0117] Figure 4 A schematic diagram of another parking space marking method provided in an embodiment of the present application;
[0118] Figure 5 A schematic flow chart of another parking space marking method provided in an embodiment of the present application;
[0119] Figure 6 A schematic diagram of a flow chart of another parking space marking method provided in an embodiment of the present application;
[0120] Figure 7 A schematic diagram of a flow chart of another parking space marking method provided in an embodiment of the present application;
[0121] Figure 8 A schematic diagram of a flow chart of another parking space marking method provided in an embodiment of the present application;
[0122] Fig. 9 A schematic diagram of a sample data collection method flow provided in an embodiment of the present application;
[0123] Fig.10 A schematic structural diagram of a parking space marking device is provided for an embodiment of the present application. DETAILED DESCRIPTION
[0124] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of this application.
[0125] Figure 1 A schematic diagram of a parking space sample data collection device provided in an embodiment of the present application, such as Figure 1 As shown, the acquisition device of this embodiment can be set on a vehicle, and the acquisition device includes: an image acquisition device, a motion data recording module, a positioning module, a processor, and the like.
[0126] The image acquisition device may be a surround fisheye camera, which is arranged around the vehicle, and includes a front fisheye camera 11, a left fisheye camera 12, a rear fisheye camera 13, and a right fisheye camera 14. The specific number of fisheye cameras is not limited in the embodiment of the present application. Figure 1Four are shown as an example, and there may be more or less. For example, the number of fisheye cameras may be determined according to the model of the vehicle or the application scenario of the vehicle. The surround fisheye camera is used to capture fisheye images. Fisheye cameras at different positions are used to capture fisheye images from different angles, so that the central computing unit can obtain a parking space image based on the fisheye images captured at different angles. The parking space image may be a bird's-eye view of the parking space.
[0127] The motion data recording module may be an inertial measurement unit (IMU) module 2, and the IMU module is used to record IMU data of vehicle driving.
[0128] The positioning module may be a Global Positioning System (GPS) module 3 , and the GPS module is used to record the GPS coordinate value of the vehicle.
[0129] The processor may be a central computing unit 4, which is used to synthesize fisheye images taken at different angles into a bird's-eye view, wherein the bird's-eye view is as follows: Figure 2 As shown, the parking space includes: corner point 0, corner point 1, corner point 2 and corner point 3, entrance line, dividing line and low edge line. Figure 2 A schematic diagram of a bird's-eye view of a parking space is provided in an embodiment of the present application, and the bird's-eye view, IMU data and GPS coordinate values are stored, thereby obtaining parking space sample data.
[0130] In the embodiment of the present application, multiple parking space images of the same parking lot are collected, some parking space images are first annotated, and then the unannotated parking space images are annotated based on the annotation information of the annotated parking space images. For the convenience of description, in the embodiment of the present application, the annotated parking space images are described as the second parking space images, and the parking space images to be annotated are described as the first parking space images. Since the first annotated parking space image has no annotation information of the annotated parking space images to be used, the first annotated parking space image can obtain the first annotation information, that is, the pixel annotation information, based on the image recognition method or the pre-annotation model, and then, the coordinate conversion matrix is used to convert the first annotated parking space image into a pixel annotation information. The second annotation information of the first annotated parking space image is obtained by using the marking method of the parking space according to the following embodiments, that is, the second annotation information of the second parking space image is obtained. Then, based on the marking method of the parking space according to the following embodiments, other unannotated first parking space images are annotated to obtain the first target annotation information of the first parking space image, and the marking of the first parking space image is completed. Then, the first parking space image is used as the second parking space image, and the second annotation information can be obtained based on the first target annotation information. Based on the second annotation information of the second parking space image, other parking space images of the same parking lot are annotated.
[0131] Figure 3 A flowchart of a parking space marking method provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the method of this embodiment is as follows:
[0132] S10: Acquire first annotation information of the first parking space image.
[0133] The first parking space image is a parking space image to be annotated, and the first annotation information is annotated based on a first coordinate system.
[0134] The first coordinate system may be a pixel coordinate system, and the first annotation information may be, for example, pixel annotation information, which refers to the coordinate positions of key points (such as corner points) of the parking space image in the pixel coordinate system. These coordinates may be based on the upper left corner of the image as the origin and in pixels to accurately mark the boundary and shape of the parking space.
[0135] The first annotation information of the first parking space image may be obtained by the following possible implementation methods:
[0136] One possible implementation is to obtain first annotation information of the first parking space image based on image recognition. For example, a convolutional neural network (CNN) in deep learning can be used to process the parking space image to identify the outline and key points (such as corner points) of the parking space.
[0137] Another possible implementation is: inputting the first parking space image into a pre-annotated model to obtain first annotation information of the first parking space image. For example, the pre-annotated model may be a machine learning model trained with a large number of images with parking space annotation information, which can identify parking space boundaries in the parking space image and output pixel coordinates of four corner points of each parking space.
[0138] S30: Acquire third annotation information of the first parking space image based on the second annotation information of the second parking space image and the first coordinate conversion matrix corresponding to the first parking space image.
[0139] The second parking space image and the first parking space image are collected in the same parking lot, the second annotation information is annotated based on a second coordinate system, the first coordinate conversion matrix is used to convert the annotation information of the second coordinate system to the first coordinate system, and the third annotation information is annotation information based on the first coordinate system.
[0140] The second coordinate system may be, for example, a terrestrial coordinate system (e.g., a WGS-84 coordinate system), and the second annotation information refers to the parking space information annotated in the terrestrial coordinate system. Such information may include geographic coordinates of the parking space, such as longitude and latitude, which provide the absolute position of the parking space in the terrestrial coordinate system.
[0141] In practical applications, the second annotation information may be obtained through a high-precision Global Positioning System (GPS) device or other Geographic Information System (GIS).
[0142] A possible implementation of this step is as follows Figure 4 As shown:
[0143] S401: Acquire second annotation information of the first parking space image based on second annotation information of the second parking space image.
[0144] Optionally, first obtain a mapping point in the first parking space image of a coordinate value corresponding to the second annotation information of the second parking space image, and then determine the second annotation information of the first parking space image based on the coordinate value of the mapping point in the first parking space image.
[0145] The second parking space image refers to a marked parking space image. The second marking information of the second parking space image is marking information based on the earth coordinate system. For example, it can be the coordinate information of the four corner points of the second parking space. The coordinate information of all corner points of each second parking space is traversed and mapped in the first parking space image to obtain the corresponding mapping points. Then, based on the coordinate values of the mapping points in the first parking space image, the second marking information of the first parking space image is determined, that is, the coordinate information of each corner point of the parking space in the first parking space image based on the earth coordinate system is determined.
[0146] Assume: There are two parking space images, namely: the first parking space image (Image1) and the second parking space image (Image2), both of which are collected in the same parking lot at different times. The second parking space image (Image2) has obtained the annotation information based on the earth coordinate system, which includes the geographical coordinates (longitude and latitude) of the four corner points of the second parking space. First, extract its annotation information from the second parking space image (Image2), assuming that this information is: the coordinate value of corner point 1: (lon21, lat21), the coordinate value of corner point 2 (lon22, lat22), the coordinate value of corner point 3 (lon23, lat23), the coordinate value of corner point 4 (lon24, lat24); based on the coordinate value corresponding to the second annotation information of the second parking space image, determine the mapping point in the earth coordinate system of the first parking space image, and based on the coordinate value of the mapping point in the earth coordinate system, the second annotation information of the first parking space image can be determined, that is, the annotation information of the first parking space image in the earth coordinate system.
[0147] Optionally, in some scenarios, for example, a corner point in the first parking space image may appear in multiple annotated second parking space images, and there may be multiple mapping points for the same corner point in the first parking space. In this scenario, the mapping points in the first parking space image may be divided into ranges, and the coordinate values of at least one mapping point in the same range may be fused to obtain multiple fused points; and the second annotation information of the first parking space image may be obtained based on the multiple fused points.
[0148] In this step, the second annotation information of the second parking space image is obtained based on the following method: acquiring the first annotation information of the second parking space image; acquiring the second annotation information of the second parking space image based on the first annotation information of the second parking space image and a second coordinate conversion matrix corresponding to the second parking space image, wherein the second coordinate conversion matrix is used to convert the annotation information of the first coordinate system into the second coordinate system.
[0149] That is, firstly, the annotation information of the second parking space image in the pixel coordinate system is acquired, and then, based on the annotation information of the second parking space image in the pixel coordinate system and the second coordinate conversion matrix corresponding to the second parking space image, the annotation information in the pixel coordinate system is converted into annotation information in the earth coordinate system, thereby obtaining the second annotation information of the second parking space image.
[0150] Optionally, in this step, the second coordinate transformation matrix can be obtained by:
[0151] Acquire fourth coordinate information of the acquisition device in the second parking space image, wherein the fourth coordinate information is coordinate information based on the first coordinate system; acquire fifth coordinate information of the acquisition device in the second parking space image, wherein the fifth coordinate information is coordinate information based on the second coordinate system; and obtain a second coordinate conversion matrix corresponding to the second parking space image based on the fourth coordinate information and the fifth coordinate information.
[0152] Among them, the acquisition device can be a collection vehicle, that is, the coordinate information of the collection vehicle that collects the second parking space image in the pixel coordinate system of the second parking space image is obtained; and the coordinate information of the collection vehicle in the earth coordinate system of the second parking space image is obtained, and based on the coordinate information in the pixel coordinate system and the coordinate information in the second coordinate system, the second coordinate conversion matrix corresponding to the second parking space image is obtained, that is, the conversion matrix from the pixel coordinate system to the earth coordinate system is obtained.
[0153] S403: Acquire third annotation information of the first parking space image based on a first coordinate conversion matrix corresponding to the first parking space image and second annotation information of the first parking space image.
[0154] The second annotation information of the first parking space image is obtained through step S401, that is, the annotation information of the first parking space image in the earth coordinate system. Combined with the first coordinate conversion matrix, the annotation information of the first parking space image in the pixel coordinate system can be obtained, that is, the third annotation information of the first parking space image is obtained.
[0155] In this step, a possible way to obtain the first coordinate transformation matrix is as follows: Figure 5 As shown:
[0156] S501: Acquire first coordinate information of a collection device in the first parking space image.
[0157] The first coordinate information is coordinate information based on the first coordinate system.
[0158] That is, first coordinate information of the acquisition device in the pixel coordinate system in the first parking space image is acquired.
[0159] S502: Acquire second coordinate information of the acquisition device in the first parking space image.
[0160] The second coordinate information is coordinate information based on the second coordinate system.
[0161] That is, the second coordinate information of the acquisition device in the earth coordinate system in the first parking space image is acquired.
[0162] The method for acquiring the second coordinate information of the acquisition device in the first parking space image is as follows: based on the correspondence between the first parking space image and the position information of the acquisition device in the second coordinate system, the second coordinate information of the acquisition device in the first parking space image is acquired, wherein the position information of the acquisition device in the second coordinate system is the coordinate information of the acquisition device in the second coordinate system when the acquisition device acquires the first parking space image.
[0163] Specifically, based on the identification information of the first parking space image, the position information of the acquisition device corresponding to the identification information in the second coordinate system can be acquired. Based on the position information of the acquisition device corresponding to the identification information in the second coordinate system, the second coordinate information of the acquisition device in the first parking space image can be determined.
[0164] That is, when the first parking space image is acquired, the position information of the acquisition device in the earth coordinate system is recorded, and a corresponding relationship between the first parking space image and the position information of the acquisition device in the earth coordinate system is established.
[0165] Optionally, the identification information may be timestamp information. The timestamp of the first parking space image is generated based on the synthesis time of the first parking space image, and the position information of the acquisition device in the earth coordinate system has a corresponding timestamp. Therefore, a correspondence between the first parking space image and the position information of the acquisition device in the earth coordinate system may be established based on the timestamp information.
[0166] More specifically, based on the identification information, the position information of the acquisition device in the third coordinate system corresponding to the identification information and the third coordinate information of the coordinate origin of the third coordinate system in the second coordinate system are obtained; based on the position information of the acquisition device in the third coordinate system and the third coordinate information of the coordinate origin of the third coordinate system in the second coordinate system, the position information of the acquisition device in the second coordinate system is obtained.
[0167] Optionally, the identification information may be timestamp information, the third coordinate system may be the acquisition device coordinate system, the coordinate origin of the acquisition device coordinate system may be the position where the acquisition device is started, and the position information of the acquisition device in the acquisition device coordinate system when the acquisition device acquires the first parking space image, as well as the coordinate information of the coordinate origin of the acquisition device in the earth coordinate system may be recorded; based on the position information of the acquisition device in the acquisition device coordinate system when the acquisition device acquires the first parking space image and the coordinate information of the coordinate origin of the acquisition device in the earth coordinate system, the position information of the acquisition device in the earth coordinate system may be deduced.
[0168] S503: Acquire a first coordinate conversion matrix corresponding to the first parking space image based on the first coordinate information and the second coordinate information.
[0169] That is, according to the position information of the acquisition device in the pixel coordinate system and the earth coordinate system at the same time, the first coordinate conversion matrix corresponding to the first parking space image can be derived, and the first coordinate conversion matrix is used to convert the coordinate value from the earth coordinate system to the pixel coordinate system.
[0170] S50: Obtaining first target annotation information of the first parking space image based on the first annotation information of the first parking space image and the third annotation information.
[0171] Optionally, a possible implementation is as follows Figure 6 As shown:
[0172] S601: Determine whether the first parking space image has the first annotation information. If yes, execute S602; otherwise, execute S605.
[0173] S602: Determine, according to a preset rule, whether the third annotation information of the first parking space image is within the first parking space image. If so, execute S603; if not, execute S604.
[0174] The preset rule may be that if the third annotation information of at least two corner points is simultaneously within the first parking space image, then it is considered to be within the first parking space image.
[0175] S603: For each corner point of the parking space, if both the first annotation information and the third annotation information exist, the first annotation information and the third annotation information are merged to obtain the first target annotation information of the corner point; if only the first annotation information exists, the first annotation information is determined to be the first target annotation information of the corner point; if only the third annotation information exists, the third annotation information is determined to be the first target annotation information of the corner point.
[0176] That is, the first annotation information is obtained based on the pre-annotated model in the first parking space image, and the third annotation information is inferred based on the annotation information of the annotated parking space image. Then, for each corner point of the parking space, the following three cases are processed:
[0177] 1) If both the first annotation information and the third annotation information exist, the first annotation information and the third annotation information are merged to obtain the first target annotation information of the corner point.
[0178] 2) If only the first annotation information exists, determining that the first annotation information is the first target annotation information of the corner point.
[0179] 3) If only the third annotation information exists, determining that the third annotation information is the first target annotation information of the corner point.
[0180] S604: According to the preset rule, if it is determined that the third annotation information of the first parking space image is not within the first parking space image, then the first annotation information of the first parking space image is determined to be the first target annotation information of the first parking space image.
[0181] S605: If the first parking space image does not have the first annotation information, determine that the third annotation information is the first target annotation information of the first parking space image.
[0182] That is, if the first parking space image does not obtain the first annotation information through the pre-annotation model, that is, the corner point of the parking space is not identified, the third annotation information inferred based on the annotation information of the annotated parking space image is determined as the first target annotation information of the first parking space image, that is, the final annotation information of the first parking space image.
[0183] In this embodiment, by acquiring first annotation information of the first parking space image, third annotation information of the first parking space image is acquired based on the second annotation information of the second parking space image and the first coordinate conversion matrix corresponding to the first parking space image, and first target annotation information of the first parking space image is obtained based on the first annotation information and the third annotation information of the first parking space image. The first annotation information and the third annotation information are annotation information obtained in two different ways. Therefore, the first target annotation information obtained by combining the first annotation information and the third annotation information has higher accuracy.
[0184] Figures 3 to 6 The illustrated embodiment is applicable to a scenario in which, in the same parking lot, the annotation information of an unlabeled parking space image is calibrated based on the annotation information of a labeled parking space image.
[0185] The embodiment of the present application also provides another scenario, that is, in the same parking lot, under different scenarios such as different time, weather, and light, the annotation of the latter scenario can be annotated based on the annotation information of the scenario already annotated in the same parking lot, so as to improve the annotation efficiency. For example, if the annotation information of each parking space image of the parking lot collected in the morning has been annotated, and it is necessary to annotate each parking space image of the parking lot collected in the afternoon, the annotation information of each parking space image of the parking lot collected in the morning can be used to annotate each parking space image of the parking lot collected in the afternoon, so as to improve the annotation efficiency. Specifically, the embodiment of this scenario is as follows:
[0186] Figure 7 A flowchart of another parking space marking method provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the method of this embodiment is as follows:
[0187] S70: Acquire fourth annotation information of the first parking space image based on the first target annotation information of the first parking space image.
[0188] The first parking space image refers to a parking space image collected in a previously annotated scene, and the first target annotation information of the first parking space image is based on Figure 3-Figure 6 The fourth marking information is the marking information based on the second coordinate system.
[0189] That is, the first target annotation information of the first parking space image is the annotation information of the pixel coordinate system obtained based on the above method, and the annotation information of the pixel coordinate system is converted to the annotation information of the earth coordinate system. The coordinate system conversion can be performed based on the coordinate conversion method described in the above embodiments.
[0190] S80: Acquire fifth annotation information of the third parking space image based on mapping points of the fourth annotation information of the first parking space image on the third parking space image.
[0191] The fifth annotation information is annotated based on the second coordinate system, wherein the third parking space image and the first parking space image are parking spaces in the same parking lot acquired under different acquisition conditions.
[0192] That is, through S70, the annotation information of the earth coordinate system of the parking space image marked in the previous marked scene of the same parking lot is obtained, and then, the annotation information is mapped to the third parking space image collected this time to obtain the corresponding mapping points, thereby obtaining the annotation information of the earth coordinate system of the third parking space image.
[0193] S90: Obtaining second target annotation information of the third parking space image based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image.
[0194] The second target labeling information is labeled based on the first coordinate system.
[0195] That is, after acquiring the annotation information of the earth coordinate system of the third parking space image, the annotation information of the earth coordinate system is converted into the annotation information of the pixel coordinate system based on the first coordinate conversion matrix.
[0196] In this embodiment, fourth annotation information of the first parking space image is obtained based on the first target annotation information of the first parking space image, fifth annotation information of the third parking space image is obtained based on the mapping point of the fourth annotation information of the first parking space image on the third parking space image, and second target annotation information of the third parking space image is obtained based on the fifth annotation information of the third parking space image and a first coordinate conversion matrix corresponding to the third parking space image. That is, when the same parking lot is annotated again, annotation is performed based on the annotation information of the previously annotated parking space images of the parking lot, thereby improving the efficiency of parking space annotation.
[0197] exist Figure 7 Based on the embodiment shown, in order to further improve the efficiency of parking space image annotation, the present application also provides a parking space annotation method, which Figure 7 The annotation information obtained in the embodiment shown is fused with the annotation information obtained by other means, such as Figure 8 As shown, Figure 8 A flowchart of another parking space marking method provided in an embodiment of the present application is shown in FIG. Figure 8 As shown, Figure 8 is Figure 7 On the basis of the embodiment shown, further, it also includes:
[0198] S100: Acquire sixth annotation information of the third parking space image.
[0199] The sixth annotation information is annotated based on the first coordinate system.
[0200] The first coordinate system may be a pixel coordinate system, and the sixth annotation information may be, for example, pixel annotation information, which refers to the coordinate positions of key points (such as corner points) of the parking space image in the pixel coordinate system. These coordinates may be based on the upper left corner of the image as the origin and in pixels to accurately mark the boundary and shape of the parking space.
[0201] The sixth annotation information of the third parking space image may be obtained by the following possible implementation manners:
[0202] One possible implementation is to obtain the sixth annotation information of the third parking space image based on image recognition. For example, a convolutional neural network (CNN) in deep learning can be used to process the parking space image to identify the outline and key points (such as corner points) of the parking space.
[0203] Another possible implementation is: inputting the third parking space image into a pre-annotated model to obtain sixth annotation information of the third parking space image. For example, the pre-annotated model may be a machine learning model trained with a large number of images with parking space annotation information, which can identify parking space boundaries in the parking space image and output pixel coordinates of four corner points of each parking space.
[0204] S110: Obtaining seventh annotation information of the third parking space image based on the fifth annotation information of the third parking space image and a first coordinate conversion matrix corresponding to the third parking space image.
[0205] The seventh annotation information is annotated based on the first coordinate system.
[0206] S120: Obtaining second target annotation information of the third parking space image based on the sixth annotation information and the seventh annotation information.
[0207] Optionally, the corner points may be divided into two categories, wherein the first category of corner points are the corner points at both ends of the parking space entrance line; and the second category of corner points are the corner points at both ends of the parking space bottom line. For the first category of corner points: the sixth annotation information and the seventh annotation information are merged to obtain the second target annotation information of the first category of corner points; for the second category of corner points: the seventh annotation information is determined to be the second target annotation information of the second category of corner points.
[0208] In this embodiment, by acquiring the sixth annotation information of the third parking space image, the seventh annotation information of the third parking space image is obtained based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image, and the second target annotation information of the third parking space image is obtained based on the sixth annotation information and the seventh annotation information. Since the sixth annotation information and the seventh annotation information are annotation information obtained based on different annotation methods, the second target annotation information of the third parking space image obtained based on the sixth annotation information and the seventh annotation information has higher accuracy.
[0209] The first parking space image, the second parking space image and the third parking space image in the above embodiments of the present application are all sample data. The embodiments of the present application also include a step of acquiring sample data, wherein the sample data includes: parking space image samples, and position information of the acquisition device in the second coordinate system when acquiring the parking space image samples, wherein the parking space image samples include: the first parking space image and the second parking space image.
[0210] Optionally, the position information of the acquisition device in the second coordinate system includes: coordinate information of the acquisition device in the second coordinate system.
[0211] Optionally, the position information of the acquisition device in the second coordinate system includes: coordinate information of the acquisition device in a third coordinate system, and coordinate information of the origin of the third coordinate system in the second coordinate system.
[0212] Optionally, the parking space image samples and the position information of the acquisition device in the second coordinate system when the parking space image samples are acquired have a corresponding relationship.
[0213] Optionally, the parking space image samples are named with identification information, and the position information of the acquisition device in the second coordinate system when acquiring the parking space image samples corresponds to the identification information.
[0214] The following describes the collection of sample data from a parking lot as an example. Fig. 9 shown.
[0215] S91: Start the data collection program.
[0216] After a vehicle equipped with data collection equipment enters the target parking lot, the data collection program can be started to collect sample data.
[0217] Among them, the data collection program can be automatically triggered to start. For example, based on the positioning information, it can be determined that the vehicle enters the parking area of the target parking lot, that is, the data collection program is automatically triggered to start. It can also be manually started. When the vehicle enters the parking area of the parking lot, the staff manually starts the data collection program. This application does not impose any restrictions on this.
[0218] S92: Record the GPS location information of the vehicle startup location.
[0219] The GPS location information of the vehicle startup position specifically refers to the GPS coordinate data of the vehicle's location when the acquisition program is started.
[0220] Optionally, the start time of the data collection program may also be recorded, and the start time of the data collection program may be used to identify the collection process of the parking space sample data.
[0221] S93: Acquire and store image information and vehicle motion status data during the vehicle's driving process.
[0222] The image information is a bird's-eye view of the parking space obtained by synthesizing the fisheye image collected by the image acquisition device during the vehicle's driving process, and the bird's-eye view is stored.
[0223] When storing the bird's-eye view, a correspondence between the bird's-eye view and the timestamp is established. For example, the bird's-eye view may be named according to the timestamp corresponding to the time when the bird's-eye view is synthesized.
[0224] Record the vehicle's motion status data, such as IMU data. Each IMU data record includes: the timestamp corresponding to the time when the IMU data was collected, the offset angle (angle) of the vehicle's current position relative to the vehicle's starting position, the horizontal and vertical coordinate information of the vehicle's current position in the vehicle's starting position coordinate system, etc.
[0225] S94: The sample data collection work is completed, and the sample data collected this time is obtained.
[0226] The sample data includes: image information of all parking spaces, vehicle motion state data, and GPS location information of the vehicle start position. In this embodiment, the sample data includes a bird's-eye view of all parking spaces, IMU data, and GPS coordinate data. The sample data is packaged as sample data to be labeled.
[0227] In this embodiment, by starting the data collection program, recording the GPS location information of the vehicle's starting position, obtaining and storing the image information of the vehicle during driving and the vehicle's motion status data, the sample data collection work is completed, and the parking space sample data collected this time is obtained, thereby preparing a data basis for the marking of the parking spaces.
[0228] Fig.10 A schematic diagram of a parking space marking device is provided in an embodiment of the present application. Fig.10As shown, the apparatus of this embodiment includes: an acquisition module 1001 and a processing module 1002, wherein the acquisition module 1001 is used to acquire first annotation information of a first parking space image, wherein the first parking space image is a parking space image to be annotated, and the first annotation information is annotated based on a first coordinate system; the processing module 1002 is used to acquire third annotation information of the first parking space image based on second annotation information of the second parking space image and a first coordinate conversion matrix corresponding to the first parking space image, wherein the second parking space image and the first parking space image are collected in the same parking lot, the second annotation information is annotated based on a second coordinate system, the first coordinate conversion matrix is used to convert the annotation information of the second coordinate system into the first coordinate system, and the third annotation information is annotation information based on the first coordinate system; first target annotation information of the first parking space image is obtained based on the first annotation information and the third annotation information of the first parking space image.
[0229] Optionally, the processing module 1002 is specifically configured to obtain second annotation information of the first parking space image based on second annotation information of the second parking space image; and obtain third annotation information of the first parking space image based on a first coordinate transformation matrix corresponding to the first parking space image and the second annotation information of the first parking space image.
[0230] Optionally, the processing module 1002 is further used to obtain first coordinate information of the acquisition device in the first parking space image, wherein the first coordinate information is coordinate information based on the first coordinate system; obtain second coordinate information of the acquisition device in the first parking space image, wherein the second coordinate information is coordinate information based on the second coordinate system; and obtain a first coordinate conversion matrix corresponding to the first parking space image based on the first coordinate information and the second coordinate information.
[0231] Optionally, the processing module 1002 is specifically used to obtain second coordinate information of the acquisition device in the first parking space image based on the correspondence between the first parking space image and the position information of the acquisition device in the second coordinate system, wherein the position information of the acquisition device in the second coordinate system is the coordinate information of the acquisition device in the second coordinate system when the acquisition device acquires the first parking space image.
[0232] Optionally, the processing module 1002 is specifically configured to obtain, based on identification information of the first parking space image, position information of the acquisition device corresponding to the identification information in the second coordinate system;
[0233] Based on the position information of the acquisition device in the second coordinate system corresponding to the identification information, second coordinate information of the acquisition device in the first parking space image is determined.
[0234] The optional processing module 1002 is specifically used to obtain, based on the identification information, the position information of the acquisition device corresponding to the identification information in the third coordinate system, and the third coordinate information of the coordinate origin of the third coordinate system in the second coordinate system;
[0235] Based on the position information of the acquisition device in the third coordinate system and the third coordinate information of the coordinate origin of the third coordinate system in the second coordinate system, the position information of the acquisition device in the second coordinate system is acquired.
[0236] Optionally, the processing module 1002 is specifically configured to obtain a mapping point in the first parking space image of a coordinate value corresponding to the second annotation information of the second parking space image;
[0237] Second annotation information of the first parking space image is determined based on the coordinate values of the mapping points in the first parking space image.
[0238] Optionally, the processing module 1002 is further configured to obtain first annotation information of the second parking space image;
[0239] Based on the first annotation information of the second parking space image and a second coordinate conversion matrix corresponding to the second parking space image, second annotation information of the second parking space image is acquired, wherein the second coordinate conversion matrix is used to convert the annotation information of the first coordinate system into the second coordinate system.
[0240] Optionally, the processing module 1002 is further configured to obtain fourth coordinate information of a collection device in the second parking space image, where the fourth coordinate information is coordinate information based on the first coordinate system;
[0241] Acquire fifth coordinate information of the acquisition device in the second parking space image, wherein the fifth coordinate information is coordinate information based on the second coordinate system;
[0242] A second coordinate conversion matrix corresponding to the second parking space image is obtained based on the fourth coordinate information and the fifth coordinate information.
[0243] Optionally, the processing module 1002 is specifically configured to determine, according to a preset rule, that the third annotation information of the first parking space image is within the first parking space image;
[0244] For each corner point of the parking space, if both the first annotation information and the third annotation information exist, the first annotation information and the third annotation information are fused to obtain the first target annotation information of the corner point; if only the first annotation information exists, the first annotation information is determined to be the first target annotation information of the corner point; if only the third annotation information exists, the third annotation information is determined to be the first target annotation information of the corner point.
[0245] Optionally, the processing module 1002 is further configured to determine, according to the preset rule, that if the third annotation information of the first parking space image is not in the first parking space image, then determine that the first annotation information of the first parking space image is the first target annotation information of the first parking space image.
[0246] Optionally, the processing module 1002 is specifically configured to determine that the third annotation information is the first target annotation information of the first parking space image if the first annotation information does not exist in the first parking space image.
[0247] Optionally, the first annotation information is obtained based on image recognition.
[0248] Optionally, the acquisition module 1001 is specifically configured to input the first parking space image into a pre-labeled model to obtain first labeling information of the first parking space image.
[0249] Optionally, the first coordinate system is a pixel coordinate system, and the second coordinate system is an earth coordinate system.
[0250] Optionally, the identification information is timestamp information.
[0251] Optionally, the processing module 1002 is specifically configured to divide the mapping points in the first parking space image into ranges, and perform fusion processing on the coordinate values of at least one mapping point in the same range to obtain a plurality of fusion points;
[0252] Second annotation information of the first parking space image is obtained based on the multiple fusion points.
[0253] Optionally, the processing module 1002 is further configured to obtain fourth annotation information of the first parking space image based on the first target annotation information of the first parking space image, wherein the fourth annotation information is annotated based on the second coordinate system;
[0254] acquiring fifth annotation information of the third parking space image based on a mapping point of fourth annotation information of the first parking space image on the third parking space image, wherein the fifth annotation information is annotated based on the second coordinate system, wherein the third parking space image and the first parking space image are parking spaces of the same parking lot collected under different collection conditions;
[0255] Based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image, second target annotation information of the third parking space image is obtained, wherein the second target annotation information is annotated based on the first coordinate system.
[0256] Optionally, the acquisition module 1001 is further used to acquire sixth annotation information of the third parking space image, wherein the sixth annotation information is annotated based on the first coordinate system;
[0257] The processing module 1002 is specifically configured to obtain seventh annotation information of the third parking space image based on fifth annotation information of the third parking space image and a first coordinate conversion matrix corresponding to the third parking space image, wherein the seventh annotation information is annotated based on the first coordinate system;
[0258] Based on the sixth labeling information and the seventh labeling information, second target labeling information of the third parking space image is obtained.
[0259] Optionally, the processing module 1002 is specifically configured to: for a first type of corner point: fuse the sixth annotation information and the seventh annotation information to obtain second target annotation information of the first type of corner point, wherein the first type of corner point is a corner point at both ends of a parking space entrance line;
[0260] For the second type of corner points: determine the seventh annotation information as the second target annotation information of the second type of corner points, wherein the second type of corner points are the corner points at both ends of the bottom edge line of the parking space.
[0261] Optionally, it also includes: an acquisition module, used to acquire sample data, the sample data including: parking space image samples, and position information of the acquisition device in the second coordinate system when acquiring the parking space image samples, wherein the parking space image samples include: the first parking space image and the second parking space image.
[0262] Optionally, the position information of the acquisition device in the second coordinate system includes: coordinate information of the acquisition device in the second coordinate system.
[0263] Optionally, the position information of the acquisition device in the second coordinate system includes: coordinate information of the acquisition device in a third coordinate system, and coordinate information of the origin of the third coordinate system in the second coordinate system.
[0264] Optionally, the parking space image samples and the position information of the acquisition device in the second coordinate system when the parking space image samples are acquired have a corresponding relationship.
[0265] Optionally, the parking space image samples are named with identification information, and the position information of the acquisition device in the second coordinate system when acquiring the parking space image samples corresponds to the identification information.
[0266] The device of this embodiment can be used to implement the technical solutions of the above-mentioned parking space marking method embodiments. Its implementation principles and technical effects are similar and will not be repeated here.
[0267] An embodiment of the present application provides an electronic device, including: a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the parking space marking method as described in any one of the first aspects are implemented.
[0268] An embodiment of the present application provides a vehicle, comprising: a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the parking space marking method as described in any one of the first aspects are implemented.
[0269] An embodiment of the present application provides a cloud server, comprising: a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the parking space marking method as described in any one of the first aspects are implemented.
[0270] An embodiment of the present application provides a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the parking space marking method as described in any one of the first aspects are implemented.
[0271] An embodiment of the present application provides a computer program product, which, when executed by a processor of a vehicle or a cloud server, implements the steps of the parking space marking method as described in any one of the first aspects.
[0272] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises one..." does not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the method and device in the embodiment of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted or combined. In addition, the features described with reference to certain examples may be combined in other examples.
[0273] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of a computer software product plus a necessary general hardware platform, and of course, can also be implemented by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, disk, CD, etc.), including several instructions to enable the terminal or network side device to execute the methods described in each embodiment of the present application.
[0274] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms of implementation methods without departing from the purpose of the present application and the scope of protection of the claims, and these implementation methods are all within the protection of the present application.
Claims
1. A parking space marking method, It is characterized in that include: Acquiring first annotation information of a first parking space image, wherein the first parking space image is a parking space image to be annotated, and the first annotation information is annotated based on a first coordinate system, and the acquiring the first annotation information of the first parking space image includes: inputting the first parking space image into a pre-annotated model to obtain the first annotation information of the first parking space image; acquiring third annotation information of the first parking space image based on second annotation information of the second parking space image and a first coordinate conversion matrix corresponding to the first parking space image, wherein the second parking space image and the first parking space image are collected in the same parking lot, the second annotation information is annotated based on a second coordinate system, the first coordinate conversion matrix is used to convert the annotation information of the second coordinate system to the first coordinate system, and the third annotation information is annotation information based on the first coordinate system; Based on the first annotation information and the third annotation information of the first parking space image, first target annotation information of the first parking space image is obtained, wherein the first target annotation information of the first parking space image is obtained based on the first annotation information and the third annotation information of the first parking space image, including: According to a preset rule, determining that the third annotation information of the first parking space image is within the first parking space image; For each corner point of the parking space, if both the first annotation information and the third annotation information exist, the first annotation information and the third annotation information are fused to obtain the first target annotation information of the corner point; if only the first annotation information exists, the first annotation information is determined to be the first target annotation information of the corner point; if only the third annotation information exists, the third annotation information is determined to be the first target annotation information of the corner point.
2. The method according to claim 1, It is characterized in that The acquiring third annotation information of the first parking space image based on the second annotation information of the second parking space image and the first coordinate conversion matrix corresponding to the first parking space image includes: acquiring second annotation information of the first parking space image based on second annotation information of the second parking space image; Based on a first coordinate conversion matrix corresponding to the first parking space image and second annotation information of the first parking space image, third annotation information of the first parking space image is acquired.
3. The method according to claim 2, It is characterized in that Before acquiring the third annotation information of the first parking space image based on the first coordinate conversion matrix corresponding to the first parking space image and the second annotation information of the first parking space image, the method further includes: Acquire first coordinate information of a collection device in the first parking space image, wherein the first coordinate information is coordinate information based on the first coordinate system; Acquire second coordinate information of the acquisition device in the first parking space image, wherein the second coordinate information is coordinate information based on the second coordinate system; A first coordinate conversion matrix corresponding to the first parking space image is acquired based on the first coordinate information and the second coordinate information.
4. The method according to claim 3, It is characterized in that The acquiring second coordinate information of the acquisition device in the first parking space image includes: Based on the correspondence between the first parking space image and the position information of the acquisition device in the second coordinate system, second coordinate information of the acquisition device in the first parking space image is acquired, wherein the position information of the acquisition device in the second coordinate system is the coordinate information of the acquisition device in the second coordinate system when the acquisition device acquires the first parking space image.
5. The method according to claim 4, It is characterized in that The acquiring second coordinate information of the acquisition device in the first parking space image based on the correspondence between the first parking space image and the position information of the acquisition device in the second coordinate system includes: Based on the identification information of the first parking space image, acquiring position information of the acquisition device corresponding to the identification information in the second coordinate system; Based on the position information of the acquisition device in the second coordinate system corresponding to the identification information, second coordinate information of the acquisition device in the first parking space image is determined.
6. The method according to claim 5, It is characterized in that The acquiring, based on the identification information of the first parking space image, the position information of the acquisition device corresponding to the identification information in the second coordinate system includes: Based on the identification information, acquiring position information of the acquisition device corresponding to the identification information in a third coordinate system, and third coordinate information of a coordinate origin of the third coordinate system in the second coordinate system; Based on the position information of the acquisition device in the third coordinate system and the third coordinate information of the coordinate origin of the third coordinate system in the second coordinate system, the position information of the acquisition device in the second coordinate system is acquired.
7. The method according to claim 2, It is characterized in that The acquiring the second annotation information of the first parking space image based on the second annotation information of the second parking space image includes: Acquire a mapping point in the first parking space image of a coordinate value corresponding to the second annotation information of the second parking space image; Second annotation information of the first parking space image is determined based on the coordinate values of the mapping points in the first parking space image.
8. The method according to claim 7, It is characterized in that The step of acquiring a coordinate value corresponding to the second annotation information of the second parking space image that is before a mapping point in the first parking space image further includes: Acquire first annotation information of the second parking space image; Based on the first annotation information of the second parking space image and a second coordinate conversion matrix corresponding to the second parking space image, second annotation information of the second parking space image is acquired, wherein the second coordinate conversion matrix is used to convert the annotation information of the first coordinate system into the second coordinate system.
9. The method according to claim 8, It is characterized in that Before acquiring the second annotation information of the second parking space image based on the first annotation information of the second parking space image and the second coordinate transformation matrix corresponding to the second parking space image, the method further includes: Acquire fourth coordinate information of a collection device in the second parking space image, where the fourth coordinate information is coordinate information based on the first coordinate system; Acquire fifth coordinate information of the acquisition device in the second parking space image, wherein the fifth coordinate information is coordinate information based on the second coordinate system; A second coordinate conversion matrix corresponding to the second parking space image is obtained based on the fourth coordinate information and the fifth coordinate information.
10. The method according to claim 1, It is characterized in that Also includes: According to the preset rule, if it is determined that the third annotation information of the first parking space image is not within the first parking space image, then the first annotation information of the first parking space image is determined to be the first target annotation information of the first parking space image.
11. The method according to any one of claims 1 to 9, It is characterized in that The obtaining, based on the first annotation information of the first parking space image and the third annotation information, first target annotation information of the first parking space image includes: If the first parking space image does not have the first annotation information, the third annotation information is determined to be the first target annotation information of the first parking space image.
12. The method according to any one of claims 1 to 9, It is characterized in that The first annotation information is obtained based on image recognition.
13. The method according to any one of claims 1 to 9, It is characterized in that The first coordinate system is a pixel coordinate system, and the second coordinate system is an earth coordinate system.
14. The method according to claim 5 or 6, It is characterized in that The identification information is timestamp information.
15. The method according to claim 7, It is characterized in that The determining, based on the coordinate values of the mapping points in the first parking space image, second annotation information of the first parking space image includes: Dividing the mapping points in the first parking space image into ranges, and fusing the coordinate values of at least one mapping point in the same range to obtain a plurality of fused points; Second annotation information of the first parking space image is obtained based on the multiple fusion points.
16. The method according to claim 1, It is characterized in that Also includes: Based on the first target annotation information of the first parking space image, acquiring fourth annotation information of the first parking space image, wherein the fourth annotation information is annotated based on the second coordinate system; acquiring fifth annotation information of the third parking space image based on a mapping point of fourth annotation information of the first parking space image on the third parking space image, wherein the fifth annotation information is annotated based on the second coordinate system, wherein the third parking space image and the first parking space image are parking spaces of the same parking lot collected under different collection conditions; Based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image, second target annotation information of the third parking space image is obtained, wherein the second target annotation information is annotated based on the first coordinate system.
17. The method according to claim 16, It is characterized in that Also includes: Acquire sixth annotation information of the third parking space image, wherein the sixth annotation information is annotated based on the first coordinate system; The obtaining the second target annotation information of the third parking space image based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image includes: Based on the fifth annotation information of the third parking space image and the first coordinate conversion matrix corresponding to the third parking space image, obtaining seventh annotation information of the third parking space image, wherein the seventh annotation information is annotated based on the first coordinate system; Based on the sixth labeling information and the seventh labeling information, second target labeling information of the third parking space image is obtained.
18. The method according to claim 17, It is characterized in that The obtaining, based on the sixth annotation information and the seventh annotation information, second target annotation information of the third parking space image includes: For the first type of corner points: the sixth annotation information and the seventh annotation information are merged to obtain second target annotation information of the first type of corner points, wherein the first type of corner points are corner points at both ends of the parking space entrance line; For the second type of corner points: determine the seventh annotation information as the second target annotation information of the second type of corner points, wherein the second type of corner points are the corner points at both ends of the bottom edge line of the parking space.
19. An electronic device, It is characterized in that include: A processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the parking space marking method according to any one of claims 1 to 18 are implemented.
20. A vehicle, It is characterized in that include: A processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the parking space marking method according to any one of claims 1 to 18 are implemented.
21. A computer-readable storage medium, It is characterized in that The readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the parking space marking method according to any one of claims 1 to 18 are implemented.
Citation Information
Patent Citations
Image processing method and device
CN114531580A
Parking space marking method and device, parking method and device, electronic equipment and storage medium
CN116486406A
Lane line labeling method, electronic device and storage medium
US20240221390A1
