Parking space identification method, electronic equipment and computer readable storage medium

Abstract

The invention discloses a parking space identification method, electronic equipment and a computer readable storage medium. The parking space identification method comprises the following steps: acquiring a target aerial view image of a target parking space; inputting the target aerial view image into a preset multi-task model, so that each task layer in the preset multi-task model outputs predicted angular point coordinate information of the target parking space, parking space type information of the target parking space, common parking space segmentation information and mechanical parking space segmentation information; and adjusting the predicted angular point coordinate information based on segmentation information corresponding to the parking space type information in the common parking space segmentation information and the mechanical parking space segmentation information. Based on the above mode, the accuracy of parking stall corner detection can be improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent driving technology, and in particular to a parking space recognition method, an electronic device, and a computer-readable storage medium. Background Art

[0002] In the prior art, an image of the parking space to be parked is usually obtained to determine the corner points of the parking space to be parked according to the image to realize parking space recognition. Subsequently, the user or the intelligent processing module can control the vehicle to park in the parking space to be parked according to the determined corner points.

[0003] The defect of the prior art is that the accuracy of corner point determination of the existing parking space recognition method for different types of parking spaces fluctuates greatly, and because the corner point determination is usually determined when the vehicle is far away from the parking space to be parked, the accuracy of the detected corner points is usually poor, and in the subsequent parking process of the vehicle, it will not be possible to obtain a complete image of all corner points of the parking space to be parked for re-determination of the corner points, which makes it easy for the corner points determined by the existing parking space recognition method to have large errors, which can easily lead to accidents in the vehicle parking operation. In summary, the accuracy of parking space corner point detection of the existing parking space recognition method is poor. Summary of the invention

[0004] The main technical problem solved by this application is how to improve the accuracy of parking space corner point detection.

[0005] In order to solve the above-mentioned technical problems, the first technical solution adopted in the present application is: a parking space recognition method, comprising: obtaining a target bird's-eye view image of a target parking space; inputting the target bird's-eye view image into a preset multi-task model, so that each task layer in the preset multi-task model respectively outputs the predicted corner point coordinate information of the target parking space, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information; based on the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information, the predicted corner point coordinate information is adjusted.

[0006] Among them, before inputting the target bird's-eye view image into the preset multi-task model, the parking space recognition method also includes: obtaining multiple parking space bird's-eye view images; respectively marking the corner point coordinates, parking space type and parking space dividing line in each parking space bird's-eye view image to obtain multiple parking space bird's-eye view image samples; and training the preset multi-task model based on the parking space bird's-eye view image samples.

[0007] Among them, the task layers in the preset multi-task model include a parking space detection task layer, an ordinary parking space segmentation task layer and a mechanical parking space segmentation task layer. The parking space detection task layer is used to output predicted corner point coordinate information and parking space type information, the ordinary parking space segmentation task layer is used to output ordinary parking space segmentation information, and the mechanical parking space segmentation task layer is used to output mechanical parking space segmentation information; the preset multi-task model is trained based on parking space bird's-eye view image samples, including: training the preset multi-task model based on parking space bird's-eye view image samples until the loss function of the parking space detection task layer, the loss function of the ordinary parking space segmentation task layer and the loss function of the mechanical parking space segmentation task layer respectively reach corresponding preset conditions.

[0008] Among them, the loss function of the parking space detection task layer is associated with the deviation of the predicted corner point coordinate information output by the preset multi-task model and the deviation of the output parking space type information; the loss function of the ordinary parking space segmentation task layer is associated with the deviation of the ordinary parking space segmentation information output by the preset multi-task model; the loss function of the mechanical parking space segmentation task layer is associated with the deviation of the mechanical parking space segmentation information output by the preset multi-task model.

[0009] Among them, based on the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information, the predicted corner point coordinate information is adjusted, including: based on the center point of the segmentation area indicated by the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information, the predicted corner point coordinate information is adjusted.

[0010] Among them, the ordinary parking space division information includes the parking space division line, and the mechanical parking space division information includes the parking space division line and the slope line.

[0011] Among them, the parking space type information includes ordinary parking spaces and mechanical parking spaces, ordinary parking spaces correspond to ordinary parking space segmentation information, and mechanical parking spaces correspond to mechanical parking space segmentation information; ordinary parking spaces include at least one of vertical parking spaces, horizontal parking spaces and inclined parking spaces.

[0012] Wherein, obtaining a parking space bird's-eye view image of a target parking space includes: obtaining at least one fisheye image of the target parking space; obtaining a target bird's-eye view image based on the fisheye image, or splicing multiple fisheye images to obtain a target bird's-eye view image.

[0013] In order to solve the above technical problems, the second technical solution adopted in the present application is: an electronic device, comprising: a memory and a processor; the memory is used to store program instructions, and the processor is used to execute the program instructions to implement the above method.

[0014] In order to solve the above technical problems, the third technical solution adopted in the present application is: a computer-readable storage medium, which stores program instructions, and the above method is implemented when the program instructions are executed by a processor.

[0015] The beneficial effect of the present application lies in that, different from the prior art, in the technical solution of the present application, a target bird's-eye view image of the target parking space is obtained, and the target bird's-eye view image is input into a preset multi-task model, so that each task layer in the preset multi-task model respectively outputs the predicted corner point coordinate information of the target parking space, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information, and the predicted corner point coordinate information is adjusted based on the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information. Based on the above method, the target bird's-eye view image can be processed based on the preset multi-task model to output the predicted corner point coordinate information corresponding to the target parking space, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information, wherein the ordinary parking space segmentation information is the segmentation information corresponding to when the target parking space is an ordinary parking space, and the mechanical parking space segmentation information is the segmentation information corresponding to when the target parking space is a mechanical parking space. When the parking space type information of the target parking space is an ordinary parking space, the predicted corner point coordinate information can be adjusted based on the ordinary parking space segmentation information, and when the parking space type information of the target parking space is a mechanical parking space, the predicted corner point coordinate information can be adjusted based on the mechanical parking space segmentation information. The predicted corner point coordinate information is adjusted according to the segmentation information. When the vehicle is approaching or entering the target parking space, the accuracy of the segmentation information will be higher than that of the predicted corner point coordinate information, and different types of parking spaces have different types of segmentation information characteristics. Therefore, based on the different types of target parking spaces, the corresponding types of segmentation information are used to specifically adjust the predicted corner point coordinate information to achieve correction, which can improve the accuracy of the predicted corner point coordinate information when the vehicle is parking in the target parking space, thereby improving the safety and reliability of the vehicle parking process as much as possible. In summary, the accuracy of parking space corner point detection is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 It is a flowchart of an embodiment of the parking space identification method of the present application;

[0018] Figure 2 It is a structural diagram of an embodiment of the preset multi-task model of the present application;

[0019] Figure 3 is a schematic diagram of an embodiment of the target parking space of the present application;

[0020] Figure 4 It is a schematic diagram of the structure of an embodiment of the electronic device of the present application;

[0021] Figure 5 It is a structural diagram of an embodiment of the computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0022] The present application is further described in detail below in conjunction with the accompanying drawings and examples. It is particularly noted that the following examples are only used to illustrate the present application, but are not intended to limit the scope of the present application. Similarly, the following examples are only some embodiments of the present application rather than all embodiments, and all other embodiments obtained by those of ordinary skill in the art without making creative work are within the scope of protection of the present application.

[0023] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0024] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "install", "set", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or connected through an intermediate medium. For ordinary technicians in this field, the above-mentioned specific meanings belonging to this application can be connected according to specific circumstances.

[0025] This application proposes a parking space recognition method, see Figure 1 , Figure 1 FIG. 1 is a flow chart of an embodiment of the parking space identification method of the present application. Figure 1 As shown, the parking space recognition method includes:

[0026] Step S11: Acquire a target bird's-eye view image of the target parking space.

[0027] Among them, at least one initial image of the target parking space can be first obtained through a camera or other types of image acquisition modules, and then the at least one initial image can be subjected to perspective conversion or other processing to obtain a top view of the target parking space, which is recorded as the target bird's-eye view image. The initial image can specifically be a fisheye image or other types of images, which are not limited here.

[0028] Step S12: Input the target bird's-eye view image into the preset multi-task model, so that each task layer in the preset multi-task model outputs the predicted corner point coordinate information of the target parking space, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information respectively.

[0029] Among them, the preset multi-task model can specifically be a pre-trained model capable of performing multiple tasks simultaneously. It can execute parking space detection tasks to output the predicted corner point coordinate information of the target parking space and the parking space type information of the target parking space, and can also execute at least two types of parking space segmentation tasks to output ordinary parking space segmentation information and mechanical parking space segmentation information.

[0030] The predicted corner point coordinate information may specifically refer to the corner points and related information of the corner point positions predicted by the corresponding task based on the target bird's-eye view image.

[0031] The parking space type information may specifically refer to relevant information about the category to which the target parking space belongs, which is predicted by the corresponding task based on the target bird's-eye view image.

[0032] The ordinary parking space segmentation information may specifically be the parking space segmentation information predicted after processing the target bird's-eye view image based on the task layer trained with samples related to ordinary parking spaces, such as a segmentation line (such as a parking space line or a lane line or other types of segmentation lines).

[0033] The mechanical parking space segmentation information may specifically be parking space segmentation information predicted after processing the target bird's-eye view image based on the task layer trained with mechanical parking space related samples, such as a segmentation line (such as a parking space line or a lane line or other types of segmentation lines).

[0034] Step S13: adjusting the predicted corner point coordinate information based on the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information.

[0035] Among them, it should be noted that since the step of obtaining the predicted corner point coordinate information is usually carried out when the vehicle is far away from the target parking space and can capture a large part of the target parking space, the relevant graphics of the four complete corner points cannot usually be obtained in the process of the subsequent vehicle parking in the target parking space. Therefore, it is impossible to determine the corner point information in real time and adjust the predicted corner point coordinate information in real time to achieve correction.

[0036] In the technical solution of the present application, based on the above method, firstly, ordinary parking space segmentation information and mechanical parking space segmentation information both belong to parking space segmentation information. Unlike the predicted corner point coordinate information, the parking space segmentation information has good generalization, that is, whether the parking space segmentation information is obtained at a distance far from or close to the target parking space, its accuracy is relatively high. Therefore, at any stage when the vehicle is parked in the target parking space, the predicted corner point coordinate information can be adjusted based on the parking space segmentation information. For example, the predicted corner point coordinate information can be dynamically adjusted based on the parking space segmentation information obtained in real time during the process of the vehicle parking in the target parking space, so as to improve the accuracy of the predicted corner point coordinate information, thereby improving the safety and reliability of the vehicle when parking in the target parking space.

[0037] Secondly, ordinary parking spaces usually refer to relatively spacious conventional parking spaces, such as parking spaces that are circled by drawing corresponding rectangular lines. Mechanical parking spaces usually refer to relatively narrow three-dimensional parking spaces, such as parking spaces that can be raised or lowered or translated after vehicles are parked in underground parking lots. The two types of parking spaces have their own characteristics. When it is detected that the target parking space is an ordinary parking space, the predicted corner point coordinate information can be adjusted by using the ordinary parking space segmentation information predicted based on the prediction method that the target parking space is an ordinary parking space. When it is detected that the target parking space is a mechanical parking space, the predicted corner point coordinate information can be adjusted by using the mechanical parking space segmentation information predicted based on the prediction method that the target parking space is a mechanical parking space. By adjusting the corner points of the target parking space of the corresponding type of parking space through the segmentation information output by the corresponding task, the accuracy of the predicted corner point coordinate information when the target parking space is an ordinary parking space and a mechanical parking space can be further improved, and the safety and reliability of the vehicle when parking in the target parking space can be further improved.

[0038] It should be noted that the parking space type information includes ordinary parking spaces and mechanical parking spaces.

[0039] Ordinary parking spaces correspond to ordinary parking space segmentation information, and mechanical parking spaces correspond to mechanical parking space segmentation information. That is, when the parking space type information is ordinary parking space, the ordinary parking space segmentation information can be used as the segmentation information of the target parking space, and when the parking space type information is mechanical parking space, the mechanical parking space segmentation information can be used as the segmentation information of the target parking space. Based on the segmentation information output by different task layers for different parking space types, the prediction of the corner point coordinate information is adjusted, which can further improve the accuracy of the prediction of the corner point coordinate information.

[0040] Ordinary parking spaces include at least one of vertical parking spaces, horizontal parking spaces, and oblique parking spaces. A vertical parking space may be a parking space where the angle between the long side of the parking space and the long side of the vehicle is 80-90 degrees, a parallel parking space may be a parking space where the angle between the long side of the parking space and the long side of the vehicle is 0-10 degrees, and an oblique parking space may be a parking space where the angle between the long side of the parking space and the long side of the vehicle is 10-80 degrees. The above are only examples, and other angle ranges may also be allocated and set, which is not limited here.

[0041] Different from the prior art, in the technical solution of the present application, a target bird's-eye view image of the target parking space is obtained, and the target bird's-eye view image is input into a preset multi-task model, so that each task layer in the preset multi-task model outputs the predicted corner point coordinate information of the target parking space, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information respectively, and the predicted corner point coordinate information is adjusted based on the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information. Based on the above method, the target bird's-eye view image can be processed based on the preset multi-task model to output the predicted corner point coordinate information, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information respectively corresponding to the target parking space, wherein the ordinary parking space segmentation information is the segmentation information corresponding to when the target parking space is an ordinary parking space, and the mechanical parking space segmentation information is the segmentation information corresponding to when the target parking space is a mechanical parking space. When the parking space type information of the target parking space is an ordinary parking space, the predicted corner point coordinate information can be adjusted based on the ordinary parking space segmentation information, and when the parking space type information of the target parking space is a mechanical parking space, the predicted corner point coordinate information can be adjusted based on the mechanical parking space segmentation information. The predicted corner point coordinate information is adjusted according to the segmentation information. When the vehicle is approaching or entering the target parking space, the accuracy of the segmentation information will be higher than that of the predicted corner point coordinate information, and different types of parking spaces have different types of segmentation information characteristics. Therefore, based on the different types of target parking spaces, the corresponding types of segmentation information are used to specifically adjust the predicted corner point coordinate information to achieve correction, which can improve the accuracy of the predicted corner point coordinate information when the vehicle is parking in the target parking space, thereby improving the safety and reliability of the vehicle parking process as much as possible. In summary, the accuracy of parking space corner point detection is improved.

[0042] In one embodiment, step S11 may specifically include:

[0043] At least one fisheye image of the target parking space is acquired.

[0044] Based on the fisheye image, a bird's-eye view image of the target is obtained, or a plurality of fisheye images are spliced ​​to obtain a bird's-eye view image of the target.

[0045] Specifically, in one example, a fisheye image of the target parking space may be acquired, and the perspective of the fisheye image may be converted to obtain an image from a top-down perspective, that is, a bird's-eye view image of the target.

[0046] In another example, multiple fisheye images of the target parking space may be acquired, and the number of the multiple fisheye images may be more than two. The multiple fisheye images are stitched and the perspective is converted to obtain an image from a top-down perspective, that is, a bird's-eye view image of the target.

[0047] Subsequently, the predicted corner point coordinate information of the target parking space can be acquired and adjusted based on the target bird's-eye view image, so as to realize the detection and identification of the target parking space according to the predicted corner point coordinate information, thereby improving the accuracy of parking space corner point detection.

[0048] In one embodiment, before step S12, the parking space recognition method further includes:

[0049] Get multiple parking space bird's-eye view images.

[0050] The corner point coordinates, parking space type and parking space dividing line in each parking space bird's-eye view image are respectively marked to obtain a plurality of parking space bird's-eye view image samples.

[0051] The preset multi-task model is trained based on parking space bird's-eye view image samples.

[0052] Specifically, a bird's-eye view image of a target parking space under various conditions can be obtained by first acquiring a fisheye image and performing a perspective conversion on the fisheye image, which is recorded as a parking space bird's-eye view image. The parking space bird's-eye view image can specifically be a bird's-eye view image of the target parking space obtained when the vehicle is in different relative position relationships with the target parking space, which is not limited here.

[0053] Manual labeling or other labeling methods can be used to label the corner point coordinates, parking space types and parking space dividing lines that can be clearly observed in multiple parking space bird's-eye views. Corner point coordinates, parking space types or parking space dividing lines that cannot be observed or whose clarity is less than a preset clarity threshold will not be labeled.

[0054] In addition, in one example, the ordinary parking space segmentation information may include a parking space segmentation line, and the mechanical parking space segmentation information may include a parking space segmentation line and a slope line. If the parking space type of the target parking space is a mechanical parking space, the location of the uphill or downhill structure in the parking space bird's-eye view image of the mechanical parking space may also be marked as a slope line, so that the preset multi-task model obtained by subsequent training based on the parking space bird's-eye view image samples with slope lines can have the ability to detect and identify the slope line of the target parking space as a mechanical parking space, so that the user can view the slope line, thereby improving the safety and reliability of parking.

[0055] Based on the above, multiple parking space bird's-eye view images are labeled separately to obtain multiple parking space bird's-eye view image samples. Subsequently, the preset multi-task model can be trained based on the parking space bird's-eye view image samples, so that the preset multi-task model can have the ability to output the predicted corner point coordinate information of the target parking space, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information through each task layer.

[0056] Alternatively, see Figure 2 , Figure 2 is a structural diagram of an embodiment of the preset multi-task model of the present application, such as Figure 2 As shown, the task layers in the preset multi-task model include a parking space detection task layer, an ordinary parking space segmentation task layer and a mechanical parking space segmentation task layer. The parking space detection task layer is used to output predicted corner point coordinate information and parking space type information, the ordinary parking space segmentation task layer is used to output ordinary parking space segmentation information, and the mechanical parking space segmentation task layer is used to output mechanical parking space segmentation information.

[0057] The preset multi-task model is trained based on the parking space bird's-eye view image samples, which may include:

[0058] The preset multi-task model is trained based on the parking space bird's-eye view image samples until the loss function of the parking space detection task layer, the loss function of the ordinary parking space segmentation task layer, and the loss function of the mechanical parking space segmentation task layer reach the corresponding preset conditions respectively.

[0059] Specifically, by merging the tasks of outputting predicted corner point coordinate information and parking space type information into the parking space detection task layer, it is possible to determine the parking space type based on a single model, distinguish between ordinary parking spaces and mechanical parking spaces, and obtain the predicted corner point coordinate information of all four corner points of the target parking space without using prior knowledge, thereby saving computing power resources and facilitating updating and maintenance.

[0060] See also Figure 3 , Figure 3 is a schematic diagram of an embodiment of the target parking space of the present application, such as Figure 3 As shown, the four corner points P1 (x1, y1), P2 (x2, y2), P3 (x3, y3) and P4 (x4, y4) of the target parking space can be the anchor point Q of any point in the target parking space.

[0061] The loss function of the parking space detection task layer is associated with the deviation of the predicted corner point coordinate information output by the preset multi-task model and the deviation of the output parking space type information.

[0062] In the parking space detection task layer, the loss function of the output parking space type information can be specifically shown as follows:

[0063]

[0064] In the parking space detection task layer, the loss function of the output predicted corner point coordinate information can be specifically shown as follows:

[0065]

[0066] Among them, that is, in the loss function of the parking space detection task layer, N represents the number of parking space bird's-eye view image samples, α and γ represent the corresponding preset weights, and y i Indicates parking space type information, p i represents the predicted probability value, Represents the predicted offset between the anchor point and the corresponding corner point in the x direction, Represents the predicted offset between the anchor point and the corresponding corner point in the y direction, Indicates the true value offset between the anchor point and the corresponding corner point in the x direction, Indicates the true value offset between the anchor point and the corresponding corner point in the y direction.

[0067] For example, Indicates the anchor point Q and the corner point P in the i-th parking space bird's-eye view image sample j The predicted offset in the x direction, Indicates the anchor point Q and the corner point P in the i-th parking space bird's-eye view image sample j The predicted offset in the y direction, and so on, the variables corresponding to the true value offset and the variables corresponding to the predicted offset are set correspondingly in meaning.

[0068] The loss function of the ordinary parking space segmentation task layer is associated with the deviation of the ordinary parking space segmentation information output by the preset multi-task model. The loss function of the mechanical parking space segmentation task layer is associated with the deviation of the mechanical parking space segmentation information output by the preset multi-task model.

[0069] In the common parking space segmentation task layer and the mechanical parking space segmentation task layer, the loss function of the output common parking space segmentation information and the mechanical parking space segmentation information can be specifically shown as follows:

[0070]

[0071] Among them, that is, in the loss function of the ordinary parking space segmentation task layer and the mechanical parking space segmentation task layer, N represents the number of parking space bird's-eye view image samples, α and γ represent the corresponding preset weights, and y i Indicates parking space type information, p i Represents the predicted probability value.

[0072] It should be noted that the parking space type information in the loss function of the ordinary parking space segmentation task layer is ordinary parking space, and the parking space type information in the loss function of the mechanical parking space segmentation task layer is mechanical parking space.

[0073] Based on the loss functions of each task layer constructed as above, it can be that in the continuous training process, when the value of the corresponding loss function reaches the preset loss function threshold, it is determined that the training of the task layer corresponding to the corresponding loss function is completed, that is, the preset conditions are met. When the loss functions corresponding to all task layers respectively reach the preset conditions, it can be determined that the training of the entire preset multi-task model is completed, and subsequently the above steps S11-S13 can be executed based on the trained preset multi-task model, thereby improving the accuracy of parking corner point detection.

[0074] In one embodiment, step S13 may specifically include:

[0075] The predicted corner point coordinate information is adjusted based on the center point of the segmented area indicated by the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information.

[0076] Specifically, after obtaining the corresponding parking space division information, based on the center point of the parking space area formed by the parking space division line in the parking space division information, it can be determined whether the positions of the four corner points in the predicted corner point coordinate information need to be corrected, and if correction is required, the positions of the corner points are adjusted based on the center point.

[0077] Since the parking space segmentation information is different from the predicted corner point coordinate information, it has good generalization, that is, the accuracy of the parking space segmentation information is high regardless of whether it is obtained at a distance far from or close to the target parking space. Therefore, the predicted corner point coordinate information can be adjusted based on the parking space segmentation information at any stage when the vehicle is parked in the target parking space. For example, the predicted corner point coordinate information can be dynamically adjusted based on the parking space segmentation information obtained in real time during the process of the vehicle parking in the target parking space, thereby improving the accuracy of the predicted corner point coordinate information, thereby improving the safety and reliability of the vehicle when parking in the target parking space.

[0078] This application also proposes an electronic device, see Figure 4 , Figure 4 is a schematic diagram of the structure of an embodiment of the electronic device of the present application, such as Figure 4 As shown, the electronic device 20 includes a processor 21 , a memory 22 and a bus 23 .

[0079] The processor 21 and the memory 22 are connected to the bus 23 respectively. The memory 22 stores program instructions. The processor 21 is used to execute the program instructions to implement the map construction method in the above embodiment.

[0080] In this embodiment, the processor 21 may also be referred to as a CPU (Central Processing Unit). The processor 21 may be an integrated circuit chip having signal processing capabilities. The processor 21 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. A general-purpose processor may be a microprocessor or the processor 21 may also be any conventional processor, etc.

[0081] Different from the prior art, in the technical solution of the present application, a target bird's-eye view image of the target parking space is obtained, and the target bird's-eye view image is input into a preset multi-task model, so that each task layer in the preset multi-task model outputs the predicted corner point coordinate information of the target parking space, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information respectively, and the predicted corner point coordinate information is adjusted based on the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information. Based on the above method, the target bird's-eye view image can be processed based on the preset multi-task model to output the predicted corner point coordinate information, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information respectively corresponding to the target parking space, wherein the ordinary parking space segmentation information is the segmentation information corresponding to when the target parking space is an ordinary parking space, and the mechanical parking space segmentation information is the segmentation information corresponding to when the target parking space is a mechanical parking space. When the parking space type information of the target parking space is an ordinary parking space, the predicted corner point coordinate information can be adjusted based on the ordinary parking space segmentation information, and when the parking space type information of the target parking space is a mechanical parking space, the predicted corner point coordinate information can be adjusted based on the mechanical parking space segmentation information. The predicted corner point coordinate information is adjusted according to the segmentation information. When the vehicle is approaching or entering the target parking space, the accuracy of the segmentation information will be higher than that of the predicted corner point coordinate information, and different types of parking spaces have different types of segmentation information characteristics. Therefore, based on the different types of target parking spaces, the corresponding types of segmentation information are used to specifically adjust the predicted corner point coordinate information to achieve correction, which can improve the accuracy of the predicted corner point coordinate information when the vehicle is parking in the target parking space, thereby improving the safety and reliability of the vehicle parking process as much as possible. In summary, the accuracy of parking space corner point detection is improved.

[0082] This application also proposes a computer-readable storage medium, see Figure 5 , Figure 5 is a schematic diagram of the structure of an embodiment of the computer-readable storage medium of the present application. Figure 5 As shown, the computer-readable storage medium 30 stores program instructions 31 thereon, and when the program instructions 31 are executed by a processor (not shown), the map construction method in the above embodiment is implemented.

[0083] The computer-readable storage medium 30 of this embodiment can be, but is not limited to, a USB flash drive, an SD card, a PD optical drive, a mobile hard disk, a large-capacity floppy drive, a flash memory, a multimedia memory card, a server, a storage unit in an FPGA or an ASIC, etc.

[0084] Different from the prior art, in the technical solution of the present application, a target bird's-eye view image of the target parking space is obtained, and the target bird's-eye view image is input into a preset multi-task model, so that each task layer in the preset multi-task model outputs the predicted corner point coordinate information of the target parking space, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information respectively, and the predicted corner point coordinate information is adjusted based on the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information. Based on the above method, the target bird's-eye view image can be processed based on the preset multi-task model to output the predicted corner point coordinate information, the parking space type information of the target parking space, the ordinary parking space segmentation information and the mechanical parking space segmentation information respectively corresponding to the target parking space, wherein the ordinary parking space segmentation information is the segmentation information corresponding to when the target parking space is an ordinary parking space, and the mechanical parking space segmentation information is the segmentation information corresponding to when the target parking space is a mechanical parking space. When the parking space type information of the target parking space is an ordinary parking space, the predicted corner point coordinate information can be adjusted based on the ordinary parking space segmentation information, and when the parking space type information of the target parking space is a mechanical parking space, the predicted corner point coordinate information can be adjusted based on the mechanical parking space segmentation information. The predicted corner point coordinate information is adjusted according to the segmentation information. When the vehicle is approaching or entering the target parking space, the accuracy of the segmentation information will be higher than that of the predicted corner point coordinate information, and different types of parking spaces have different types of segmentation information characteristics. Therefore, based on the different types of target parking spaces, the corresponding types of segmentation information are used to specifically adjust the predicted corner point coordinate information to achieve correction, which can improve the accuracy of the predicted corner point coordinate information when the vehicle is parking in the target parking space, thereby improving the safety and reliability of the vehicle parking process as much as possible. In summary, the accuracy of parking space corner point detection is improved.

[0085] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradicting each other.

[0086] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0087] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0088] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (which can be a personal computer, server, network device or other system that can fetch instructions from the instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0089] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A parking space recognition method, characterized in that: include: Acquire a target bird's-eye view image of a target parking space; Inputting the target bird's-eye view image into a preset multi-task model, so that each task layer in the preset multi-task model outputs predicted corner point coordinate information of the target parking space, parking space type information of the target parking space, ordinary parking space segmentation information, and mechanical parking space segmentation information respectively; The predicted corner point coordinate information is adjusted based on segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information.

2. The parking space recognition method according to claim 1, characterized in that: Before inputting the target bird's-eye view image into a preset multi-task model, the parking space recognition method further includes: Obtaining bird's-eye views of multiple parking spaces; Respectively marking the corner point coordinates, parking space type and parking space dividing line in each parking space bird's-eye view image to obtain a plurality of parking space bird's-eye view image samples; The preset multi-task model is trained based on the parking space bird's-eye view image samples.

3. The parking space recognition method according to claim 2, characterized in that: The task layers in the preset multi-task model include a parking space detection task layer, a common parking space segmentation task layer and a mechanical parking space segmentation task layer, wherein the parking space detection task layer is used to output the predicted corner point coordinate information and the parking space type information, the common parking space segmentation task layer is used to output the common parking space segmentation information, and the mechanical parking space segmentation task layer is used to output the mechanical parking space segmentation information; The training of the preset multi-task model based on the parking space bird's-eye view image sample includes: The preset multi-task model is trained based on the parking space bird's-eye view image samples until the loss function of the parking space detection task layer, the loss function of the ordinary parking space segmentation task layer and the loss function of the mechanical parking space segmentation task layer respectively reach corresponding preset conditions.

4. The parking space recognition method according to claim 3, characterized in that: The loss function of the parking space detection task layer is associated with the deviation of the predicted corner point coordinate information output by the preset multi-task model and the deviation of the parking space type information output; The loss function of the common parking space segmentation task layer is associated with the deviation degree of the common parking space segmentation information output by the preset multi-task model; The loss function of the mechanical parking space segmentation task layer is associated with the deviation of the mechanical parking space segmentation information output by the preset multi-task model.

5. The parking space recognition method according to claim 1, characterized in that: The adjusting the predicted corner point coordinate information based on the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information includes: The predicted corner point coordinate information is adjusted based on the center point of the segmented area indicated by the segmentation information corresponding to the parking space type information in the ordinary parking space segmentation information and the mechanical parking space segmentation information.

6. The parking space recognition method according to claim 1, characterized in that: The ordinary parking space segmentation information includes a parking space segmentation line, and the mechanical parking space segmentation information includes a parking space segmentation line and a slope line.

7. The parking space recognition method according to claim 1, characterized in that: The parking space type information includes ordinary parking spaces and mechanical parking spaces, the ordinary parking spaces correspond to the ordinary parking space segmentation information, and the mechanical parking spaces correspond to the mechanical parking space segmentation information; The ordinary parking spaces include at least one of vertical parking spaces, horizontal parking spaces and inclined parking spaces.

8. The parking space recognition method according to claim 1, characterized in that: The step of obtaining a parking space bird's-eye view image of a target parking space includes: Acquire at least one fisheye image of the target parking space; The target bird's-eye view image is obtained based on the fisheye image, or the target bird's-eye view image is obtained by splicing a plurality of the fisheye images.

9. An electronic device, characterized in that: include: Memory and processor; The memory is used to store program instructions, and the processor is used to execute the program instructions to implement the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images