A parking space number identification method and device, electronic equipment and storage medium
By using panoramic cameras and multi-model recognition technology, the problems of slow speed and low accuracy of traditional parking space number recognition have been solved, achieving efficient and accurate parking space number detection and management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HANGZHOU ZOHO INFORMATION TECH CO LTD
- Filing Date
- 2022-10-14
- Publication Date
- 2026-06-16
AI Technical Summary
Traditional parking space number recognition methods are slow and have low accuracy, making them ineffective in managing the matching of vehicles and parking spaces in cities.
A panoramic camera is used to capture panoramic images of the parking lot. Through a trained parking space number location recognition model, text direction correction model, and text recognition model, the location and key point information of the parking space number are identified, and image conversion and angle correction are performed to improve recognition accuracy.
This significantly improves the speed and accuracy of parking space number detection, enhancing the efficiency and precision of parking space number management.
Smart Images
Figure CN115512285B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image processing technology, and in particular to a method, apparatus, electronic device and storage medium for identifying parking space numbers. Background Technology
[0002] In recent years, due to the expansion of urban areas, the number of vehicles in cities has also exploded. Currently, parking spaces in parking lots are usually equipped with corresponding parking space numbers. However, since the number of vehicles far exceeds the number of parking spaces, car owners often face the problem of not being able to find a parking space. Therefore, the management of automatically identifying parking space numbers and marking them as parking spaces is very important. However, the traditional method of manually processing parking space numbers has technical problems such as slow detection speed and low accuracy. Summary of the Invention
[0003] In view of this, the purpose of this application is to provide a method, device, electronic device and storage medium for identifying parking space numbers, thereby improving the speed and accuracy of parking space number detection.
[0004] This application provides a method for identifying parking space numbers, the method comprising:
[0005] The panoramic image of the parking lot to be inspected is input into the trained parking space number location recognition model to determine the location information of the parking space number in the panoramic image and the key point information of the parking space number.
[0006] The panoramic image is converted into a corresponding cubeface texture image, and the first parking space number image is determined from the cubeface texture image based on the location information of the parking space number and the key point information of the parking space number in the converted cubeface texture image.
[0007] The first parking space number image is input into the trained text direction correction model to determine the second parking space number image after the correction direction is determined.
[0008] The second parking space number image is input into the trained text recognition model to determine the parking space number information in the parking lot to be detected.
[0009] Furthermore, the trained parking space number location recognition model includes a feature extractor, a location detector, and a key point detector. The step of inputting a panoramic image of the parking lot to be inspected into the trained parking space number location recognition model to determine the location information of the parking space number and the key point information of the parking space number in the panoramic image includes:
[0010] The panoramic image of the parking lot to be inspected is input into the feature extractor in the trained parking space number location recognition model to determine the parking space number feature map corresponding to the panoramic image.
[0011] The parking space number feature map is input into the location detector in the trained parking space number location recognition model to determine the location information of the parking space number corresponding to the parking space number feature map.
[0012] The location information of the parking space number is input into the key point detector in the trained parking space number location recognition model to determine the key point information of the parking space number corresponding to the parking space number feature map.
[0013] Furthermore, the key information of the parking space number includes the information of the four corner points of the parking space number and the information of the center point of the parking space number.
[0014] Furthermore, determining the first parking space number image from the cube face map image based on the location information of the parking space number and the key point information of the parking space number in the converted cube face map image includes:
[0015] Based on the position information of the parking space number in the converted cube face texture image, the initial text image of the parking space number is determined;
[0016] Based on the key point information of the parking space number in the converted cube face texture image, the first parking space number image is determined from the cube face texture image.
[0017] Furthermore, the trained text orientation correction model is determined in the following way:
[0018] Obtain multiple sample parking space number images and sample deformation angle labels in each of the sample parking space number images;
[0019] Each of the sample parking space number images is input into the initial text direction correction model to obtain the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images.
[0020] Based on the entropy loss between the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images and the sample deformation angle label in each of the sample parking space number images, the initial text direction correction model is trained to determine the trained text direction correction model; wherein, the entropy loss is obtained by normalizing both the predicted angle value and the sample deformation angle label according to the angle period.
[0021] This application embodiment also provides a parking space number identification device, the parking space number identification device comprising:
[0022] The first determining module is used to input the panoramic image of the parking lot to be inspected into the trained parking space number location recognition model to determine the location information of the parking space number in the panoramic image and the key point information of the parking space number.
[0023] The second determining module is used to convert the panoramic image into a corresponding cube face map image, and determine the first parking space number image from the cube face map image based on the location information of the parking space number and the key point information of the parking space number in the converted cube face map image.
[0024] The third determining module is used to input the first parking space number image into the trained text direction correction model to determine the second parking space number image after correction direction.
[0025] The fourth determining module is used to input the second parking space number image into the trained text recognition model to determine the parking space number information in the parking lot to be detected.
[0026] Furthermore, the trained parking space number location recognition model includes a feature extractor, a location detector, and a key point detector. The first determining module is specifically used for:
[0027] The panoramic image of the parking lot to be inspected is input into the feature extractor in the trained parking space number location recognition model to determine the parking space number feature map corresponding to the panoramic image.
[0028] The parking space number feature map is input into the location detector in the trained parking space number location recognition model to determine the location information of the parking space number corresponding to the parking space number feature map.
[0029] The location information of the parking space number is input into the key point detector in the trained parking space number location recognition model to determine the key point information of the parking space number corresponding to the parking space number feature map.
[0030] The key information of the parking space number includes the information of the four corner points of the parking space number and the information of the center point of the parking space number.
[0031] This application embodiment also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the parking space number identification method described above are performed.
[0032] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the parking space number identification method described above.
[0033] Compared with existing parking space number recognition methods, the parking space number recognition method, electronic device, and storage medium provided in this application significantly improve the acquisition efficiency by acquiring panoramic images of the parking lot to be inspected using a panoramic camera. Furthermore, through image conversion, a trained parking space number location recognition model, a trained text direction correction model, and a trained text recognition model, the embodiments of this application can more accurately identify the parking space number information in the panoramic image of the parking lot to be inspected, thereby improving the parking space number detection speed and accuracy.
[0034] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description
[0035] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0036] Figure 1 A flowchart of one of the parking space number identification methods provided in this application embodiment is shown;
[0037] Figure 2 A second flowchart of a parking space number identification method provided in an embodiment of this application is shown;
[0038] Figure 3 A schematic diagram of the structure of a parking space number identification device provided in an embodiment of this application is shown;
[0039] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown.
[0040] In the picture:
[0041] 300 - Parking space number identification device; 310 - First determination module; 320 - Second determination module; 330 - Third determination module; 340 - Fourth determination module; 400 - Electronic device; 410 - Processor; 420 - Memory; 430 - Bus. Detailed Implementation
[0042] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.
[0043] First, the applicable application scenarios of this application will be introduced. This application can be applied to the field of image processing technology. Research has found that in recent years, due to the expansion of urban scale, the number of vehicles in cities has also exploded. Currently, parking spaces in parking lots are usually equipped with corresponding parking space numbers. However, since the number of vehicles far exceeds the number of parking spaces, car owners often face the problem of not being able to find a parking space. Therefore, the management of automatically identifying parking space numbers and marking them as parking spaces is very important. However, the traditional method of manually processing parking space numbers has technical problems of slow detection speed and low accuracy.
[0044] Based on this, embodiments of this application provide a method, apparatus, electronic device, and storage medium for identifying parking space numbers, thereby improving the speed and accuracy of parking space number detection.
[0045] Please see Figure 1 , Figure 1 This is a flowchart illustrating a parking space number identification method provided in an embodiment of this application. Figure 1 As shown in the figure, the parking space number identification method provided in this application embodiment includes the following steps:
[0046] S101. Input the panoramic image of the parking lot to be inspected into the trained parking space number location recognition model to determine the location information of the parking space number in the panoramic image and the key point information of the parking space number.
[0047] In this step, a large number of panoramic images of the parking lot to be inspected are collected using a panoramic camera. For each panoramic image of the parking lot to be inspected collected by the panoramic camera, the panoramic image of the parking lot to be inspected is input into a trained parking space number location recognition model to determine the location information of the parking space number in the panoramic image and the key point information of the parking space number.
[0048] Here, the panoramic image is used to represent the bird's-eye view of the parking lot to be inspected, which is captured by the panoramic camera. The bird's-eye view will cover as many driving scenarios as possible in the parking lot to be inspected. In the embodiments provided in this application, it is assumed that the panoramic camera captures about 20 panoramic images in each driving scenario in the parking lot to be inspected.
[0049] The pre-trained parking space number location recognition model can be customized according to different needs. For example, the pre-trained parking space number location recognition model in the embodiments provided in this application can be, but is not limited to, a pano-YOLO network.
[0050] In this way, the Pano-YOLO network treats object detection as a regression problem. After one inference pass through the input image, it can obtain the location of all objects in the image, their category, and their corresponding confidence probability.
[0051] The trained parking space number location recognition model described above is determined through the following steps:
[0052] Obtain multiple sample panoramic images and the corresponding label for each sample panoramic image.
[0053] The sample panoramic image is input into the initial parking space number location recognition model to obtain the location information of the sample parking space number in each sample panoramic image and the predicted value of the key point information of the sample parking space number.
[0054] Based on the predicted values of the sample parking space number's location information and key point information in each sample panoramic image, and the entropy loss between the corresponding label of each sample panoramic image, the initial parking space number location recognition model is trained to determine the trained parking space number location recognition model.
[0055] Optionally, the trained parking space number location recognition model includes a feature extractor, a location detector, and a key point detector. Therefore, S101 includes the following sub-steps:
[0056] Sub-step 1011: Input the panoramic image of the parking lot to be inspected into the feature extractor in the trained parking space number location recognition model to determine the parking space number feature map corresponding to the panoramic image.
[0057] In this step, the panoramic image of the parking lot to be inspected is first input into the feature extractor of the trained parking space number location recognition model. This is used to extract all the feature information related to the parking space number from the panoramic image of the parking lot to be inspected, and to determine the feature information of the parking space number as the corresponding parking space number feature map.
[0058] Sub-step 1012: Input the parking space number feature map into the position detector in the trained parking space number position recognition model to determine the position information of the parking space number corresponding to the parking space number feature map.
[0059] In this step, after determining the parking space number feature map, the parking space number feature map is input into the position detector in the trained parking space number position recognition model to detect the position information of the parking space number in the parking space number feature map and determine the position information of the parking space number corresponding to the parking space number feature map.
[0060] Sub-step 1013: Input the location information of the parking space number into the key point detector in the trained parking space number location recognition model to determine the key point information of the parking space number corresponding to the parking space number feature map.
[0061] In this step, since the panoramic image of the parking lot to be inspected (i.e., a bird's-eye view of the parking lot to be inspected) is acquired by a panoramic camera, the panoramic image may be greatly distorted in some usage scenarios of the parking lot to be inspected, resulting in a large error with the actual parking space number at that location. Therefore, the pre-trained parking space number location recognition model provided in this application has been structurally improved. In the implementation provided in this application, a key point detector is added to the pre-trained parking space number location recognition model to detect the key points of the parking space number in the panoramic image. Moreover, for the application scenario provided in the embodiment of this application, the embodiment of this application selects the information of the four corner points and the information of the midpoint of the long side of the parking space number as the key point information of the parking space number.
[0062] Here, after adding the key point detector to the trained parking space number location recognition model, the location information of the parking space number is input into the key point detector in the trained parking space number location recognition model to determine the key point information of the parking space number corresponding to the parking space number feature map.
[0063] In the process of training the parking space number location recognition model, the loss training of the key point detector can be, but is not limited to, using the mean square error loss calculation method.
[0064] Optionally, the key information of the parking space number includes the information of the four corner points of the parking space number and the information of the center point of the parking space number.
[0065] Here, the key information of the parking space number can be customized and adjusted according to different application scenarios and needs.
[0066] Specifically, it is assumed that the key point information of the parking space number in the embodiments provided in this application includes the information of the four corner points of the parking space number and the information of the midpoint of the parking space number; or the key point information of the parking space number in the embodiments provided in this application may also include the information of the four corner points of the parking space number and the information of the two midpoints of the parking space number.
[0067] S102. Convert the panoramic image into a corresponding cube face map image, and determine the first parking space number image from the cube face map image based on the location information of the parking space number and the key point information of the parking space number in the converted cube face map image.
[0068] In this step, due to the distortion and different text orientations in the panoramic image, directly using it in a trained text recognition model would severely affect accuracy. Therefore, the embodiment provided in this application, after determining the location information of the parking space number and the key point information of the parking space number in the panoramic image, converts the panoramic image into a corresponding cube facet image, then transforms it from the panoramic camera coordinate system to the planar coordinate system, determines the location information of the parking space number and the key point information of the parking space number in the converted cube facet image, and determines the first parking space number image based on the location information of the parking space number and the key point information of the parking space number in the converted cube facet image.
[0069] S103. Input the first parking space number image into the trained text direction correction model to determine the second parking space number image after correction.
[0070] In this step, after determining the first parking space number image, the first parking space number image is input into the trained text orientation correction model, and the deformation angle of the first parking space number image is output. After determining the deformation angle of the first parking space number image, the deformation angle of the first parking space number image is corrected.
[0071] Here, the trained text orientation correction model can be customized according to different use cases and different needs. It is assumed that the trained text orientation correction model in the embodiments provided in this application can be, but is not limited to, using a residual network (ResNet).
[0072] As mentioned above, residual networks are characterized by their ease of optimization and the ability to improve accuracy by increasing their depth. The residual blocks within them use skip connections, which alleviates the gradient vanishing problem caused by increasing the depth in deep neural networks.
[0073] Optionally, the trained text orientation correction model can be determined through the following sub-steps:
[0074] Sub-step 1031: Obtain multiple sample parking space number images and sample deformation angle labels in each of the sample parking space number images.
[0075] In this step, the deformation angle of each sample parking space number image in the multiple sample parking space number images can be specific but not limited to 0-360 degrees.
[0076] Here, 0 degrees represents the normal orientation, and the orientation of the sample parking space number image is rotated counterclockwise to 360 degrees.
[0077] The multiple sample parking space number images can be expanded through random online augmentation, rotation, translation, and scaling, among other things. When rotation is performed, the labels of the corresponding sample parking space number images are updated. For example, if the sample parking space number image needs to be rotated counterclockwise by 'a' degrees, then the label of the sample parking space number image needs to be increased by 'a' degrees.
[0078] In the above, the training of the text direction correction model adopts a cosine return learning rate adjustment strategy, and is specifically trained 200 times.
[0079] Sub-step 1032: Input each of the sample parking space number images into the initial text direction correction model to obtain the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images.
[0080] Sub-step 1033: Based on the entropy loss between the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images and the sample deformation angle label in each of the sample parking space number images, train the initial text direction correction model to determine the trained text direction correction model; wherein, the entropy loss is obtained by normalizing both the predicted angle value and the sample deformation angle label according to the angle period.
[0081] In this step, the formula for calculating entropy loss can be customized according to different needs or application scenarios. For example, the formula for calculating entropy loss in the embodiments provided in this application is as follows:
[0082] L=|x′-cosθ|+|y′-sinθ|;
[0083] in,
[0084]
[0085] Here, the optimizer with entropy loss is Adam, and the learning rate is initialized to 1e-3.
[0086] The entropy loss is calculated because the predicted angle value and the sample deformation angle label have periodicity. That is, 0 degrees and 360 degrees represent the same direction, but the labels for 0 degrees and 360 degrees are different. Therefore, during the training process, the predicted angle value and the sample deformation angle label are normalized according to the angle period, so that the trained text direction correction model outputs two values x and y. Since the sum of the squares of cosθ and sinθ is 1, x and y need to be normalized as well.
[0087] S104. Input the second parking space number image into the trained text recognition model to determine the parking space number information in the parking lot to be detected.
[0088] In this step, after determining the second parking space number image, the second parking space number image is input into the trained text recognition model to determine the parking space number information in the parking lot to be detected. After determining the parking space number information in the parking lot to be detected, the parking space numbers of the empty parking spaces are scheduled for vehicles to park in the parking lot to be detected, and the non-empty parking spaces are marked to remind vehicles that want to park in the parking lot to be detected.
[0089] Compared with existing parking space number recognition methods, the parking space number recognition method provided in this application significantly improves the acquisition efficiency by acquiring panoramic images of the parking lot to be inspected using a panoramic camera. Furthermore, through image conversion, a trained parking space number location recognition model, a trained text direction correction model, and a trained text recognition model, the parking space number information in the panoramic image of the parking lot to be inspected can be more accurately identified, thereby improving the parking space number detection speed and accuracy.
[0090] Please see Figure 2 , Figure 2 This is a flowchart of a second method for identifying parking space numbers according to this application. Figure 2 As shown in the figure, the parking space number identification method provided in this application embodiment includes the following steps:
[0091] S201. Input the panoramic image of the parking lot to be inspected into the trained parking space number location recognition model to determine the location information of the parking space number in the panoramic image and the key point information of the parking space number.
[0092] The trained parking space number location recognition model includes a feature extractor, a location detector, and a key point detector. The step of inputting a panoramic image of the parking lot to be inspected into the trained parking space number location recognition model to determine the location information of the parking space number and the key point information of the parking space number in the panoramic image includes:
[0093] The panoramic image of the parking lot to be inspected is input into the feature extractor in the trained parking space number location recognition model to determine the parking space number feature map corresponding to the panoramic image.
[0094] The parking space number feature map is input into the location detector in the trained parking space number location recognition model to determine the location information of the parking space number corresponding to the parking space number feature map.
[0095] The location information of the parking space number is input into the key point detector in the trained parking space number location recognition model to determine the key point information of the parking space number corresponding to the parking space number feature map.
[0096] The key information of the parking space number includes the information of the four corner points of the parking space number and the information of the center point of the parking space number.
[0097] S202. Determine the initial text image of the parking space number based on the position information of the parking space number in the converted cube face texture image.
[0098] In this step, based on the location information of the parking space number in the converted cube face map image, the area of the parking space number is cropped out on the cube face map image, and the image of this area is determined as the initial text image of the parking space number.
[0099] Here, the size of the area where the parking space number is cut out can be determined according to different usage scenarios and the accuracy requirements under different usage scenarios.
[0100] Since the coordinate systems of the panoramic image and the cube face map image are different, there is a mapping relationship between them, which can be specifically expressed as: the coordinates of the cube face map image = F (the coordinates of the panoramic image). Therefore, the initial text image of the parking space number can be directly obtained based on the coordinates of the panoramic image.
[0101] S203. Based on the key point information of the parking space number in the converted cube face texture image, determine the first parking space number image from the cube face texture image.
[0102] In this step, based on the key point information of the parking space number in the converted cube face texture image, the initial text image is further cropped according to the corresponding mapping relationship, thereby determining the first parking space number image from the cube face texture image.
[0103] S204. Input the first parking space number image into the trained text direction correction model to determine the second parking space number image after correction.
[0104] The trained text orientation correction model is determined using the following method:
[0105] Obtain multiple sample parking space number images and sample deformation angle labels in each of the sample parking space number images.
[0106] Each of the sample parking space number images is input into the initial text direction correction model to obtain the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images.
[0107] Based on the entropy loss between the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images and the sample deformation angle label in each of the sample parking space number images, the initial text direction correction model is trained to determine the trained text direction correction model; wherein, the entropy loss is obtained by normalizing both the predicted angle value and the sample deformation angle label according to the angle period.
[0108] S205. Input the second parking space number image into the trained text recognition model to determine the parking space number information in the parking lot to be detected.
[0109] The descriptions of S201 and S204 to S205 can be referred to the descriptions of S101 and S103 to S104, and can achieve the same technical effect, so they will not be elaborated further.
[0110] Compared with existing parking space number recognition methods, the parking space number recognition method provided in this application significantly improves the acquisition efficiency by acquiring panoramic images of the parking lot to be inspected using a panoramic camera. Furthermore, through image conversion, a trained parking space number location recognition model, a trained text direction correction model, and a trained text recognition model, the parking space number information in the panoramic image of the parking lot to be inspected can be more accurately identified, thereby improving the parking space number detection speed and accuracy.
[0111] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a parking space number identification device provided in an embodiment of this application. Figure 3 As shown, the parking space number identification device 300 includes:
[0112] The first determining module 310 is used to input the panoramic image of the parking lot to be inspected into the trained parking space number location recognition model to determine the location information of the parking space number in the panoramic image and the key point information of the parking space number.
[0113] Optionally, the trained parking space number location recognition model includes a feature extractor, a location detector, and a key point detector. The first determining module 310 is specifically used for:
[0114] The panoramic image of the parking lot to be inspected is input into the feature extractor in the trained parking space number location recognition model to determine the parking space number feature map corresponding to the panoramic image.
[0115] The parking space number feature map is input into the location detector in the trained parking space number location recognition model to determine the location information of the parking space number corresponding to the parking space number feature map.
[0116] The location information of the parking space number is input into the key point detector in the trained parking space number location recognition model to determine the key point information of the parking space number corresponding to the parking space number feature map.
[0117] Optionally, the key information of the parking space number includes the information of the four corner points of the parking space number and the information of the center point of the parking space number.
[0118] The second determining module 320 is used to convert the panoramic image into a corresponding cube facet image, and to determine the first parking space number image from the cube facet image based on the location information of the parking space number and the key point information of the parking space number in the converted cube facet image.
[0119] Optionally, the second determining module 320 is specifically used for:
[0120] Based on the location information of the parking space number in the converted cube face texture image, the initial text image of the parking space number is determined.
[0121] Based on the key point information of the parking space number in the converted cube face texture image, the first parking space number image is determined from the cube face texture image.
[0122] The third determining module 330 is used to input the first parking space number image into the trained text direction correction model to determine the second parking space number image after correction direction.
[0123] Optionally, the trained text orientation correction model can be determined in the following ways:
[0124] Obtain multiple sample parking space number images and sample deformation angle labels in each of the sample parking space number images.
[0125] Each of the sample parking space number images is input into the initial text direction correction model to obtain the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images.
[0126] Based on the entropy loss between the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images and the sample deformation angle label in each of the sample parking space number images, the initial text direction correction model is trained to determine the trained text direction correction model; wherein, the entropy loss is obtained by normalizing both the predicted angle value and the sample deformation angle label according to the angle period.
[0127] The fourth determining module 340 is used to input the second parking space number image into the trained text recognition model to determine the parking space number information in the parking lot to be detected.
[0128] Compared with existing parking space number recognition devices, the parking space number recognition device 300 provided in this application significantly improves the acquisition efficiency by acquiring panoramic images of the parking lot to be inspected through a panoramic camera. Furthermore, through image conversion, a trained parking space number position recognition model, a trained text direction correction model, and a trained text recognition model, the parking space number information in the panoramic image of the parking lot to be inspected can be more accurately identified, thereby improving the parking space number detection speed and accuracy.
[0129] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device 400 includes a processor 410, a memory 420, and a bus 430.
[0130] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, they can perform the operations described above. Figure 1 as well as Figure 2 The steps of the parking space number identification method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.
[0131] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 as well as Figure 2 The steps of the parking space number identification method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.
[0132] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0133] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.
[0134] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0135] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0136] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0137] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for identifying parking space numbers, characterized in that, The method for identifying the parking space number includes: The panoramic image of the parking lot to be detected is input into the trained parking space number location recognition model to determine the location information of the parking space number in the panoramic image and the key point information of the parking space number. The panoramic image is converted into a corresponding cubeface texture image, and the location information of the parking space number and the key point information of the parking space number are used as the basis for the converted cubeface texture image. Specifically, based on the position information of the parking space number in the converted cube face texture image, the initial text image of the parking space number is determined; based on the key point information of the parking space number in the converted cube face texture image, the first parking space number image is determined from the initial text image. The first parking space number image is input into the trained text direction correction model to determine the second parking space number image after the correction direction is determined. The second parking space number image is input into the trained text recognition model to determine the parking space number information in the parking lot to be detected.
2. The parking space number identification method according to claim 1, characterized in that, The trained parking space number location recognition model includes a feature extractor, a location detector, and a key point detector. The step of inputting a panoramic image of the parking lot to be detected into the trained parking space number location recognition model to determine the location information of the parking space number and the key point information of the parking space number in the panoramic image includes: The panoramic image of the parking lot to be detected is input into the feature extractor in the trained parking space number location recognition model to determine the parking space number feature map corresponding to the panoramic image. The parking space number feature map is input into the location detector in the trained parking space number location recognition model to determine the location information of the parking space number corresponding to the parking space number feature map. The location information of the parking space number is input into the key point detector in the trained parking space number location recognition model to determine the key point information of the parking space number corresponding to the parking space number feature map.
3. The parking space number identification method according to claim 1, characterized in that, The key information of the parking space number includes the information of the four corner points of the parking space number and the information of the center point of the parking space number.
4. The parking space number identification method according to claim 1, characterized in that, The trained text orientation correction model is determined using the following method: Obtain multiple sample parking space number images and sample deformation angle labels in each of the sample parking space number images; Each of the sample parking space number images is input into the initial text direction correction model to obtain the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images. Based on the entropy loss between the predicted angle value corresponding to the sample deformation angle in each of the sample parking space number images and the sample deformation angle label in each of the sample parking space number images, the initial text direction correction model is trained to determine the trained text direction correction model; wherein, the entropy loss is obtained by normalizing both the predicted angle value and the sample deformation angle label according to the angle period.
5. A parking space number identification device, characterized in that, The parking space number identification device includes: The first determining module is used to input the panoramic image of the parking lot to be detected into the trained parking space number location recognition model to determine the location information of the parking space number in the panoramic image and the key point information of the parking space number. The second determining module is used to convert the panoramic image into a corresponding cubeface texture image, and based on the position information of the parking space number and the key point information of the parking space number in the converted cubeface texture image; wherein, based on the position information of the parking space number in the converted cubeface texture image, an initial text image of the parking space number is determined; and based on the key point information of the parking space number in the converted cubeface texture image, a first parking space number image is determined from the initial text image; The third determining module is used to input the first parking space number image into the trained text direction correction model to determine the second parking space number image after correction direction. The fourth determining module is used to input the second parking space number image into the trained text recognition model to determine the parking space number information in the parking lot to be detected.
6. The parking space number identification device according to claim 5, characterized in that, The trained parking space number location recognition model includes a feature extractor, a location detector, and a key point detector. The first determining module is specifically used for: The panoramic image of the parking lot to be detected is input into the feature extractor in the trained parking space number location recognition model to determine the parking space number feature map corresponding to the panoramic image. The parking space number feature map is input into the location detector in the trained parking space number location recognition model to determine the location information of the parking space number corresponding to the parking space number feature map. The location information of the parking space number is input into the key point detector in the trained parking space number location recognition model to determine the key point information of the parking space number corresponding to the parking space number feature map.
7. The parking space number identification device according to claim 6, characterized in that, The key information of the parking space number includes the information of the four corner points of the parking space number and the information of the center point of the parking space number.
8. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the parking space number identification method as described in any one of claims 1-4.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the parking space number identification method as described in any one of claims 1-4.