Parking space identification method and device, electronic equipment and storage medium

Abstract

The invention discloses a parking space identification method and device, electronic equipment and a storage medium, and the method comprises the steps: inputting a collected parking lot image into a network model for image key point detection, and obtaining a parking space corner point position set, at least one virtual center point position and the deviation of a virtual center point relative to a corner point; and determining the position of the corner point on the parking space to which each virtual center point position belongs by utilizing the deviation of the virtual center point relative to the corner point and the parking space corner point position set. As the virtual center point is the midpoint between the midpoint of the entrance side of the parking space and the gravity center point of the parking space and is a point which is located in the middle of the parking space and is close to the vehicle, the influence of shielding and unclear images is small, and one parking space can be defined every time one virtual center point is obtained. Therefore, the accurate position of the angular point on each parking space can be determined by utilizing the deviation of the virtual central point relative to the angular point and the angular point position set, the parking space detection precision is improved, and the parking requirement is met after the vehicle is automatically parked in the parking space.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and specifically to a parking space recognition method, device, electronic device and storage medium. Background Art

[0002] In the scenario of autonomous driving parking, detecting parking spaces in the parking lot is a key task to achieve automatic parking. At present, when identifying parking spaces, the target detection method is often used to obtain the positions of the four corner points on the parking space detection frame. However, considering that the parking task has high requirements for the accuracy of the corner points of the parking space, it is difficult for this conventional target detection method to detect the corner points with the required accuracy, resulting in a large deviation between the detected parking space and the actual position. As a result, after the vehicle automatically parks in the parking space, the deviation between the vehicle position and the parking space is relatively large, which cannot meet the parking requirements. Summary of the invention

[0003] The purpose of the present application is to propose a parking space identification method, device, electronic device and storage medium to address the deficiencies of the above-mentioned prior art, and this purpose is achieved through the following technical solutions.

[0004] The first aspect of the present application provides a parking space identification method, the method comprising:

[0005] Input the collected parking lot image into a preset neural network model to detect image key points, and obtain a set of parking space corner point positions, at least one virtual center point position, and an offset of the virtual center point relative to the corner point; wherein the virtual center point is the midpoint between the midpoint of the parking space entrance edge and the parking space center of gravity;

[0006] The positions of the corner points on the parking spaces to which the positions of the virtual center points belong are determined by using the offset of the virtual center points relative to the corner points and the position set of the parking space corner points.

[0007] In some embodiments of the present application, the neural network model includes a feature extraction network, a neck network, a corner point recognition network and a parking space recognition network; the step of inputting the collected parking lot image into a preset neural network model for image key point detection includes:

[0008] The feature extraction network is used to extract image features of different scales of the parking lot image; the neck network is used to perform multi-scale feature fusion on the image features to obtain fused features; the corner point recognition network is used to perform parking space corner point detection based on the fused features to obtain a parking space corner point position set; the parking space recognition network is used to obtain at least one virtual center point position and the offset of the virtual center point relative to the corner point based on the fused features.

[0009] In some embodiments of the present application, the corner point recognition network includes a first key point detection network and a first output layer; and obtaining a parking space corner point position set based on the fusion feature through the corner point recognition network includes:

[0010] The first key point detection network extracts the thermal regression map features and position deviation features of the parking corner points according to the fusion features; the first output layer obtains multiple parking corner point positions according to the thermal regression map features, and the deviation of each parking corner point position is corrected according to the position deviation features to form a parking corner point position set.

[0011] In some embodiments of the present application, the parking space recognition network includes a second key point detection network and a second output layer; the obtaining at least one virtual center point position and an offset of the virtual center point relative to the corner point based on the fusion feature through the parking space recognition network includes:

[0012] The second key point detection network extracts the thermal regression map features of the virtual center point of the parking space and the offset features of the virtual center point relative to the corner point according to the fusion features; the second output layer obtains at least one virtual center point position according to the thermal regression map features, and obtains the feature value of the position corresponding to each virtual center point position on the offset feature as the offset of the virtual center point relative to the corner point.

[0013] In some embodiments of the present application, determining the position of the corner point on the parking space to which each virtual center point belongs by using the offset of the virtual center point relative to the corner point and the parking space corner point position set includes:

[0014] The positions of the corner points on the parking spaces to which the positions of the virtual center points belong are determined according to the offset of the virtual center points relative to the corner points; and the positions of the corner points on the parking spaces are corrected according to the parking space corner point position set.

[0015] In some embodiments of the present application, determining the position of the corner point on the parking space to which each virtual center point belongs according to the offset of the virtual center point relative to the corner point includes:

[0016] For each virtual center point position, the positions of the four types of corner points on the parking space to which the virtual center point position belongs are determined respectively by utilizing the offsets of the virtual center point corresponding to the virtual center point position relative to the four types of corner points; wherein the offsets of the virtual center point relative to the four types of corner points include the horizontal offset value and the vertical offset value of the virtual center point relative to each type of corner point.

[0017] In some embodiments of the present application, the parking space corner point position set includes four types of parking space corner point position sets; and the correction of the position of the corner point on each parking space according to the parking space corner point position set includes:

[0018] For the position of each type of corner point on each parking space, obtain the parking space corner point position matching the position in the corresponding type of parking space corner point position set, and use the matched parking space corner point position to replace the position.

[0019] In some embodiments of the present application, the step of obtaining a parking space corner point position matching the position in the corresponding parking space corner point position set includes:

[0020] The parking space corner point position closest to the position is obtained from the corresponding parking space corner point position set as the matching parking space corner point position.

[0021] A second aspect of the present application provides a parking space recognition device, the device comprising:

[0022] A model processing module is used to input the collected parking lot image into a preset neural network model to perform image key point detection, and obtain a parking space corner point position set, at least one virtual center point position, and an offset of the virtual center point relative to the corner point; wherein the virtual center point is the midpoint between the midpoint of the parking space entrance edge and the parking space center of gravity;

[0023] The parking space recognition module is used to determine the position of the corner point on the parking space to which each virtual center point belongs by using the offset of the virtual center point relative to the corner point and the parking space corner point position set.

[0024] The third aspect of the present application proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect above.

[0025] A fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement the method described in the first aspect above.

[0026] Based on the parking space recognition method and device described in the first and second aspects above, the present application has at least the following beneficial effects or advantages:

[0027] From the perspective of improving the accuracy of corner points, the parking lot image is input into the neural network model for image key point detection to obtain a relatively high-precision parking space corner point position set, at least one virtual center point position, and the offset of the virtual center point relative to the corner point. Since the virtual center point is the midpoint between the midpoint of the parking space entrance edge and the parking space center of gravity, it is a position point located in the middle of the parking space and close to the vehicle. Therefore, it is less affected by occlusion and unclear images. In this way, each virtual center point can clearly identify a parking space, which can improve the parking space detection rate. Therefore, using the offset of the virtual center point relative to the corner point and the obtained parking space corner point position set is also conducive to determining the precise position of the corner point on each parking space, improving the parking space detection accuracy, so that after the vehicle automatically parks into the parking space, the vehicle position coincides with the parking space, meeting the parking requirements.

[0028] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0030] Figure 1 The present invention is a flow chart of an embodiment of a parking space identification method according to an exemplary embodiment;

[0031] Figure 2 is a structural schematic diagram of a neural network model according to an exemplary embodiment;

[0032] Figure 3 is a schematic diagram of parking space corner point classification according to an exemplary embodiment;

[0033] Figure 4 is a schematic structural diagram of a parking space recognition device according to an exemplary embodiment;

[0034] Figure 5 is a schematic diagram of a hardware structure of an electronic device according to an exemplary embodiment;

[0035] Figure 6 The figure is a schematic diagram of the structure of a storage medium according to an exemplary embodiment. DETAILED DESCRIPTION

[0036] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0037] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0038] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0039] As mentioned above, in an existing technology before the present application, a deep learning-based target detection model is used to predict parking space detection frames on AVM (Around View Monitor) images. Each parking space detection frame consists of 4 corner points. The target detection model adopts a traditional detection model, such as a yolo model or a centernet model. The traditional detection model pays more attention to the regression of the detection frame and uses IOU (Intersection Over Union) as a measurement indicator. The position accuracy of the corner points given by the detection frame cannot meet the parking requirements.

[0040] In another prior art prior to the present application, a deep learning network is used to perform pixel-level segmentation on parking space lines on an AVM image, and then the corner point position information of the parking space is obtained through post-processing and combined with the general size of the parking space. This method has relatively high requirements for post-processing, and the edge contour of the parking space detected by segmentation is often rough, and the imaging of the AVM image will further aggravate the roughness of the detected contour. Therefore, the accuracy of the corner point position obtained from the rough contour cannot meet the parking requirements.

[0041] In view of this, the present application proposes a parking space recognition method from the perspective of improving the accuracy of corner points. The parking lot image is input into a neural network model for image key point detection to obtain a relatively high-precision parking space corner point position set, at least one virtual center point position, and the offset of the virtual center point relative to the corner point. Since the virtual center point is the midpoint between the midpoint of the parking space entrance edge and the parking space center of gravity, it is a position point located in the middle of the parking space and close to the vehicle. Therefore, it is less affected by occlusion and unclear images. In this way, each virtual center point can clearly identify a parking space, which can improve the parking space detection rate. Therefore, using the offset of the virtual center point relative to the corner point and the obtained parking space corner point position set is also conducive to determining the precise position of the corner point on each parking space, improving the parking space detection accuracy, so that after the vehicle automatically parks into the parking space, the vehicle position coincides with the parking space, meeting the parking requirements.

[0042] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The several specific embodiments listed can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0043] Figure 1 The flowchart of a parking space identification method according to an exemplary embodiment includes the following steps:

[0044] Step 101: Input the collected parking lot image into a preset neural network model to perform image key point detection to obtain a parking space corner point position set, at least one virtual center point position, and an offset of the virtual center point relative to the corner point.

[0045] In this step, the parking lot image is generally an AVM image, which can be specifically obtained by stitching together a 4-way fisheye camera. The parking space corner point position set is the corner point set on all parking spaces in the image. The virtual center point is the midpoint between the midpoint of the parking space entrance edge and the parking space center of gravity. The parking space corner point position set and the offset of the virtual center point relative to the corner point are used to accurately locate and detect the positions of the four corner points on the parking space. Considering the principle of image imaging, the closer the object is to the vehicle, the higher the recognition degree in the image. Therefore, this application detects the position point (that is, the virtual center point) located in the middle of the parking space and close to the vehicle to detect the parking space, so as to reduce the impact of unclear images and occlusions and improve the parking space detection rate.

[0046] In an alternative embodiment, if Figure 2 The neural network model structure shown in the figure includes a feature extraction network, a neck network, a corner point recognition network, and a parking space recognition network. The image key point detection process includes:

[0047] First, image features of different scales of the parking lot image are extracted through a feature extraction network. Optionally, the feature extraction network may include but is not limited to Res Net, Reg Net, Efficient Net, etc.

[0048] Then, the neck network is used to perform multi-scale feature fusion on the image features to obtain fused features. Since different scale features contain different information, large-scale features contain more corner location information, and small-scale features contain more parking space semantic information, multi-scale feature fusion is used to improve feature expression capabilities. Optionally, the neck network may include but is not limited to FPN (Feature Pyramid Network), DLA (Deep Layer Aggregation), etc.

[0049] Then, the corner point detection of the parking space is performed based on the fusion features through the corner point recognition network to obtain the parking space corner point position set. It should be noted here that the parking space corner point position set can include four types of parking space corner point position sets K1, K2, K3, and K4. The definitions of these four types of corner points are as follows: Figure 3 As shown in the figure, the right entry point in the direction of exit is taken as the starting point, and the counterclockwise order is 0, 1, 2, and 3, and four types of corner points are obtained. The line connecting the first two corner points is the parking space entrance edge, the short edge is perpendicular to the parking space entrance edge, and the long edge is horizontal to the parking space entrance edge. It can be seen that the corner point recognition network can estimate the key points of corner points according to these four categories.

[0050] Finally, the parking space recognition network obtains at least one virtual center point position and the offset of the virtual center point relative to the corner point based on the fusion features.

[0051] Depend on Figure 2 It can be seen that the corner point recognition network and the parking space recognition network are two parallel branch networks, so the two do not distinguish the execution order. Optionally, the corner point recognition network and the parking space recognition network can both be composed of convolutional layers, downsampling layers, and upsampling layers, but the layer combinations and network parameters of the two networks are different to achieve different prediction functions.

[0052] In a specific embodiment, Figure 2 As shown, the corner point recognition network includes a first key point detection network and a first output layer. The processing process for the corner point recognition network includes: extracting the thermal regression map features and position deviation features of the parking space corner points according to the fusion features through the first key point detection network, and then obtaining multiple parking space corner point positions according to the thermal regression map features through the first output layer, and performing deviation correction on each parking space corner point position according to the position deviation feature to form a parking space corner point position set.

[0053] Among them, by finding the position with the largest eigenvalue in the thermal regression map features and decreasing the eigenvalue from this position to the surrounding area, such a position is used as the parking space corner point position. Specifically, the first output layer can use NMS (Non-Maximum Suppression), threshold filtering and other operations to obtain the parking space corner point position. In order to reduce the pixel error caused by pixel rounding based on the heat map, the position deviation feature of the parking space corner point is obtained by offset regression of the parking space corner point, which is used to correct the parking space corner point extracted from the thermal regression map features, thereby obtaining a parking space corner point position set with relatively high position accuracy.

[0054] It can be understood that the method of locating corner key points based on heat map form given above is only a specific example. Under the premise of meeting the corner point position accuracy requirements, other key point estimation methods can also be used to achieve corner point positioning, such as using binary classification to locate parking space corner points.

[0055] In a specific embodiment, Figure 2 As shown, the parking space recognition network includes a second key point detection network and a second output layer. The processing process for the parking space recognition network includes: extracting the thermal regression map features of the virtual center point of the parking space and the offset features of the virtual center point relative to the corner point through the second key point detection network according to the fusion features, and then obtaining at least one virtual center point position according to the thermal regression map features through the second output layer, and obtaining the feature value corresponding to each virtual center point position on the offset feature as the offset of the virtual center point relative to the corner point.

[0056] Among them, for the process of extracting the virtual center point position according to the thermal regression map features, please refer to the aforementioned process of extracting the parking space corner points according to the thermal regression map features, and no further description is given. A virtual center point position represents a parking space, and the offset feature of the virtual center point relative to the corner point represents the association between the virtual center point and the corner points on the parking space to which it belongs. It can be seen that the processing process of the parking space recognition network is equivalent to modeling the association between the corner points on the parking space, so that the position of each corner point on the parking space can also be located by the offset of the virtual center point relative to the corner point and the position of the virtual center point, but the corner point position located in this way is less accurate than the corner point position located by the above-mentioned corner point key point method.

[0057] Based on the above Figure 2 The neural network model shown below takes the definition of four types of corner points as an example to illustrate the various features obtained during the processing of the corner point recognition network and the parking space recognition network:

[0058] The four channels of the thermal regression map feature M*N*4 of the parking corner point output by the first key point detection network in the corner point recognition network respectively represent the features of the four types of corner points, and the eight channels of the output position deviation feature M*N*8, in which every two channels represent the position deviation features of one type of corner point, and one channel out of every two channels represents the deviation of the horizontal position coordinates, and the other channel represents the deviation of the vertical position coordinates.

[0059] One channel of the thermal regression map feature M*N*1 of the virtual center point of the parking space output by the second key point detection network in the parking space recognition network represents the feature of the virtual center point, and each of the eight channels of the output offset feature M*N*8 of the virtual center point relative to the corner point represents the horizontal position offset and the vertical position offset of the virtual center point relative to a type of corner point.

[0060] Step 102: using the offset of the virtual center point relative to the corner point and the parking space corner point position set, determine the position of the corner point on the parking space to which each virtual center point position belongs.

[0061] As mentioned above, the offset of the virtual center point relative to the corner point includes the offset of each virtual center point relative to the four types of corner points. The positions of the four corner points on the parking space can also be located by the offset of the virtual center point relative to the corner point and the position of the virtual center point. However, the corner point position accuracy obtained in this way is lower than the position accuracy in the parking space corner point position set.

[0062] Based on this, in order to improve the accuracy of the corner point position, the position of the corner point on the parking space to which each virtual center point belongs can be determined according to the offset of the virtual center point relative to the corner point, and then the position of the corner point on each parking space can be corrected according to the parking space corner point position set.

[0063] As mentioned above, since the offset of the virtual center point relative to the corner point is the offset of the virtual center point relative to the four types of corner points, it includes the offset value of the virtual center point relative to each type of corner point in the horizontal direction and the offset value in the vertical direction.

[0064] Based on this, the process of determining the position of the corner point on the parking space to which each virtual center point belongs can be to determine the positions of the four types of corner points on the parking space to which the virtual center point belongs, using the offset of the virtual center point corresponding to the virtual center point position relative to the four types of corner points.

[0065] Furthermore, as mentioned above, the parking space corner point position set includes 4 types of parking space corner point position sets. Therefore, in order to match the corner point positions in the 4 types of sets to the corresponding parking spaces to achieve the correction of the corner point positions, it is possible to obtain the parking space corner point position that matches the position in the corresponding type of parking space corner point position set for the position of each type of corner point on each parking space, and replace the position with the matched parking space corner point position to achieve position correction.

[0066] Specifically, during matching, the parking space corner point position closest to the position may be obtained from the corresponding parking space corner point position set as the matching parking space corner point position.

[0067] For example, assuming that the four types of corner point position sets are K1, K2, K3, and K4, the offset of the virtual center point relative to the four types of corner points is used to obtain n groups of combinations D1, D2, D3...Dn describing the parking space corner points. Each combination corresponds to a parking space, which contains the positions of the four types of corner points. For the positions of the four types of corner points in each combination, the nearest distance matching strategy is used to match the nearest corner point positions in the four sets K1, K2, K3, and K4 to replace them.

[0068] At this point, the above Figure 1 The parking space recognition process shown in the figure starts from the consideration of improving the accuracy of corner points. By inputting the parking lot image into the neural network model for image key point detection, a relatively high-precision parking space corner point position set, at least one virtual center point position, and the offset of the virtual center point relative to the corner point are obtained. Since the virtual center point is the midpoint between the midpoint of the parking space entrance edge and the parking space center of gravity, it is a position point located in the middle of the parking space and close to the vehicle. Therefore, it is less affected by occlusion and unclear images. In this way, each virtual center point can clearly identify a parking space, which can improve the parking space detection rate. Therefore, using the offset of the virtual center point relative to the corner point and the obtained parking space corner point position set is also conducive to determining the precise position of the corner point on each parking space, improving the parking space detection accuracy, so that after the vehicle automatically parks into the parking space, the vehicle position coincides with the parking space, meeting the parking requirements.

[0069] The execution subject of the embodiment of the present application may be an application, service, instance, software-based functional module, virtual machine (VM), container or cloud server, etc., or a hardware device (such as a server or terminal device) or hardware chip (such as CPU, GPU, FPGA, NPU, AI acceleration card or DPU) with data processing function. The device for realizing parking space recognition can be deployed on the computing device of the application party providing the corresponding service or on the cloud computing platform providing computing power, storage and network resources. The mode of providing services to the outside world by the cloud computing platform may be IaaS (Infrastructure as a Service), PaaS (Platform as a Service), SaaS (Software as a Service) or DaaS (Data as a Service). Taking the platform providing SaaS software as a service (Software as a Service) as an example, the cloud computing platform can use its own computing resources to provide the training of the parking space recognition model or the functional execution of the parking space recognition module, and the specific application architecture can be built according to the service requirements. For example, the platform can provide construction services based on the above model to applications or individuals using platform resources, and further call the above model and implement online or offline parking space identification functions based on parking space identification requests submitted by relevant clients or servers and other devices.

[0070] Corresponding to the above-mentioned embodiment of the parking space recognition method, the present application also provides an embodiment of a parking space recognition device.

[0071] Figure 4 FIG. 1 is a schematic diagram showing a parking space recognition device according to an exemplary embodiment. The device is used to execute the parking space recognition method provided in any of the above embodiments. Figure 4 As shown, the parking space recognition device includes:

[0072] The model processing module 410 is used to input the collected parking lot image into a preset neural network model to perform image key point detection, and obtain a parking space corner point position set, at least one virtual center point position, and an offset of the virtual center point relative to the corner point; wherein the virtual center point is the midpoint between the midpoint of the parking space entrance edge and the parking space center of gravity;

[0073] The parking space identification module 420 is used to determine the position of the corner point on the parking space to which each virtual center point belongs by using the offset of the virtual center point relative to the corner point and the parking space corner point position set.

[0074] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0075] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0076] The embodiments of the present application also provide an electronic device corresponding to the parking space recognition method provided in the aforementioned embodiments, so as to execute the aforementioned parking space recognition method.

[0077] Figure 5 6 is a hardware structure diagram of an electronic device according to an exemplary embodiment, the electronic device includes: a communication interface 601, a processor 602, a memory 603 and a bus 604; wherein the communication interface 601, the processor 602 and the memory 603 communicate with each other through the bus 604. The processor 602 can execute the parking space recognition method described above by reading and executing the machine executable instructions corresponding to the control logic of the parking space recognition method in the memory 603. The specific content of the method is referred to the above embodiment, which will not be repeated here.

[0078] The memory 603 mentioned in this application can be any electronic, magnetic, optical or other physical storage device, and can contain storage information, such as executable instructions, data, etc. Specifically, the memory 603 can be RAM (Random Access Memory), flash memory, storage drive (such as hard disk drive), any type of storage disk (such as optical disk, DVD, etc.), or similar storage medium, or a combination thereof. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 601 (which can be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0079] The bus 604 may be an ISA bus, a PCI bus or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 603 is used to store programs, and the processor 602 executes the programs after receiving execution instructions.

[0080] Processor 602 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in processor 602 or an instruction in software form. The above-mentioned processor 602 can be a general-purpose processor, including a network processor (Network Processor, referred to as NP), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a readily available programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The disclosed methods, steps and logic block diagrams in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as a hardware decoding processor to execute, or the hardware and software modules in the decoding processor are combined to execute.

[0081] The electronic device provided in the embodiment of the present application and the parking space identification method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented therein.

[0082] The present application also provides a computer-readable storage medium corresponding to the parking space identification method provided in the above embodiment. Figure 6 As shown, the computer-readable storage medium shown is a CD 30 on which a computer program (ie, a program product) is stored. When the computer program is executed by a processor, the parking space recognition method provided by any of the aforementioned embodiments will be executed.

[0083] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.

[0084] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the parking space identification method provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the method adopted, run or implemented by the application program stored therein.

[0085] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0086] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0087] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A parking space recognition method, characterized in that: The method comprises: Input the collected parking lot image into a preset neural network model to detect image key points, and obtain a set of parking space corner point positions, at least one virtual center point position, and an offset of the virtual center point relative to the corner point; wherein the virtual center point is the midpoint between the midpoint of the parking space entrance edge and the parking space center of gravity; The positions of the corner points on the parking spaces to which the positions of the virtual center points belong are determined by using the offset of the virtual center points relative to the corner points and the position set of the parking space corner points.

2. The method according to claim 1, characterized in that The neural network model includes a feature extraction network, a neck network, a corner point recognition network and a parking space recognition network; The method of inputting the collected parking lot image into a preset neural network model to detect image key points includes: Extracting image features of different scales of the parking lot image through the feature extraction network; Performing multi-scale feature fusion on the image features through the neck network to obtain fused features; Performing parking space corner point detection based on the fusion features through the corner point recognition network to obtain a parking space corner point position set; At least one virtual center point position and an offset of the virtual center point relative to the corner point are obtained through the parking space recognition network based on the fusion feature.

3. The method according to claim 2, characterized in that The corner point recognition network includes a first key point detection network and a first output layer; The obtaining of a parking space corner point position set based on the fusion features through the corner point recognition network includes: Extracting thermal regression map features and position deviation features of parking space corner points according to the fusion features through the first key point detection network; A plurality of parking space corner point positions are obtained according to the thermal regression map features through the first output layer, and deviation correction is performed on each parking space corner point position according to the position deviation features to form a parking space corner point position set.

4. The method according to claim 2, characterized in that: The parking space recognition network includes a second key point detection network and a second output layer; The obtaining, by the parking space recognition network based on the fusion feature, at least one virtual center point position and an offset of the virtual center point relative to the corner point comprises: Extracting the thermal regression map features of the virtual center point of the parking space and the offset features of the virtual center point relative to the corner point according to the fusion features through the second key point detection network; At least one virtual center point position is obtained according to the thermal regression map feature through the second output layer, and the feature value of the position corresponding to each virtual center point position is obtained on the offset feature as the offset of the virtual center point relative to the corner point.

5. The method according to claim 1, characterized in that The method of determining the position of the corner point on the parking space to which each virtual center point belongs by using the offset of the virtual center point relative to the corner point and the parking space corner point position set comprises: Determine the position of the corner point on the parking space to which each virtual center point belongs according to the offset of the virtual center point relative to the corner point; The positions of the corner points on each parking space are corrected according to the parking space corner point position set.

6. The method according to claim 5, characterized in that The step of determining the position of the corner point on the parking space to which each virtual center point belongs according to the offset of the virtual center point relative to the corner point includes: For each virtual center point position, using the offset of the virtual center point corresponding to the virtual center point position relative to the four types of corner points, the positions of the four types of corner points on the parking space to which the virtual center point position belongs are determined respectively; The offset of the virtual center point relative to the four types of corner points includes an offset value of the virtual center point relative to each type of corner point in the horizontal direction and an offset value in the vertical direction.

7. The method according to claim 6, characterized in that The parking space corner point position set includes 4 types of parking space corner point position sets; The step of correcting the position of the corner points on each parking space according to the parking space corner point position set includes: For each type of corner point position on each parking space, obtain a parking space corner point position matching the position in the corresponding type of parking space corner point position set, and replace the position with the matched parking space corner point position; The acquiring a parking space corner point position matching the position in the corresponding parking space corner point position set includes: acquiring a parking space corner point position closest to the position in the corresponding parking space corner point position set as the matching parking space corner point position.

8. A parking space recognition device, characterized in that: The device comprises: A model processing module is used to input the collected parking lot image into a preset neural network model to perform image key point detection, and obtain a parking space corner point position set, at least one virtual center point position, and an offset of the virtual center point relative to the corner point; wherein the virtual center point is the midpoint between the midpoint of the parking space entrance edge and the parking space center of gravity; The parking space recognition module is used to determine the position of the corner point on the parking space to which each virtual center point belongs by using the offset of the virtual center point relative to the corner point and the parking space corner point position set.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes the program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the method according to any one of claims 1 to 7.

Images