A parking lot vehicle identification method, system, device and storage medium

Abstract

The present application provides a method, system, device and storage medium for vehicle identification in a parking lot, which obtains a vehicle perspective image set of a target vehicle in a hospital parking lot; performs distortion correction on the vehicle perspective image set to obtain a vehicle perspective correction image set; determines the color sudden change block of the selected vehicle perspective correction image, extracts the recognition kernel based on the convolution model, and then determines the recognition cost through the recognition kernel; obtains the image illumination information of the target vehicle, determines the amount of shooting feature loss during shooting based on the image illumination information and the recognition cost, determines the identification features of the target vehicle based on the pixels with illumination defects and the amount of shooting feature loss on the vehicle perspective image; determines whether the target vehicle is parked in a standardized manner based on the identification features and the preset identification features. Based on the identification features, the above scheme can accurately identify the vehicle contour and the position of the parking space border under the interference of complex color features, thereby standardizing the parking status of the vehicle based on the identification results.

Description

Technical Field

[0001] The present application relates to the technical field of vehicle identification, and more specifically, to a method, system, device and storage medium for vehicle identification in a parking lot. Background Art

[0002] Vehicle identification covers multiple application areas such as vehicle identity recognition, behavior monitoring, and safety management. For example, the identification of vehicles for standardized parking. Among them, the technology of vehicle standardized parking identification plays a key role in intelligent transportation systems, involving multiple fields such as computer vision, machine learning, image processing, and sensor fusion. The core technology includes collecting parking lot images through cameras, and using deep learning models such as convolutional neural networks (CNN) for license plate recognition and vehicle detection to accurately determine whether the vehicle is parked in a standardized manner. Sensor fusion technology integrates data from cameras, lidar, and ultrasonic sensors to provide more accurate vehicle location and posture information, improving system robustness and accuracy. In addition, combined with the global positioning system (GPS) and geo-fencing technology, dynamic management of parking areas can be achieved. The vehicle standard parking sign system can not only detect illegal parking, but also optimize the allocation of parking resources and improve the utilization rate and management efficiency of parking lots. The continuous optimization and innovation of these technologies have promoted the development of vehicle standard parking signs towards automation and intelligence, and provided solid technical support for the construction of smart cities. In hospital parking lots, the problem of irregular parking of vehicles is common, resulting in a large number of parking spaces being occupied, which seriously wastes precious parking resources. For example, some drivers fail to park their vehicles accurately in a single parking space when parking, and often cross the parking space line, occupying two or more parking spaces, and some vehicles are parked in no-parking areas, emergency passages, entrances and exits, etc., which hinder traffic flow and affect the passage of emergency vehicles. Therefore, it is particularly important to mark irregularly parked vehicles in hospital parking lots.

[0003] In the prior art, the implementation of standardized parking of vehicles is based on computer vision and sensor technology. First, real-time image data in the parking lot is collected by a high-resolution camera and transmitted to the central processing system for license plate recognition and vehicle detection to determine the vehicle position and posture, and then the determined vehicle position and posture are compared with the preset standard to determine whether the parking is standardized, so as to achieve intelligent and efficient parking management; however, in the identification of irregular parking of vehicles in hospital parking lots, there are relatively complex color features at the border of the parking space and the outline of the vehicle, that is, the border of the parking space will be blurred after a lot of wear and tear, and there is an edge color mixing effect between the vehicle outline color and the background color (that is, the phenomenon that the vehicle outline color and the background color are mixed together), so that the existing vehicle identification system of the hospital parking lot cannot accurately identify the position of the vehicle outline and the parking space border, resulting in the inability to standardize the parking status of the target vehicle, resulting in random parking of vehicles, affecting the overall operation efficiency and patient experience of the hospital parking lot. Therefore, how to accurately identify the position of the vehicle outline and the parking space border under the interference of complex color features, so as to standardize the parking status of the vehicle based on the identification results has become a difficult problem faced by the industry. Summary of the invention

[0004] The present application provides a method, system, device and storage medium for vehicle identification in a parking lot, which can accurately identify a target vehicle under the interference of complex color features, thereby improving the effectiveness of suspicious identification of licensed vehicles based on the identification results.

[0005] In a first aspect, the present application provides a method for identifying vehicles in a parking lot, comprising the following steps:

[0006] Obtain a vehicle-view image set of a target vehicle in a hospital parking lot;

[0007] Performing distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain a vehicle perspective correction image set;

[0008] Selecting a vehicle perspective adjustment image from the vehicle perspective adjustment image set as a selected vehicle perspective adjustment image, determining a color sudden change block corresponding to the selected vehicle perspective adjustment image, extracting an identification kernel for color recognition of the selected vehicle perspective adjustment image from the color sudden change block based on a pre-trained convolution model, continuing to determine identification kernels for color recognition of the remaining vehicle perspective adjustment images, and then determining the identification cost of the target vehicle under the image color sudden change through all the identification kernels;

[0009] Acquire image illumination information of the target vehicle, determine the amount of shooting feature loss when shooting the target vehicle under the influence of illumination according to the image illumination information and the recognition cost, and determine the identification features of the target vehicle based on all pixel points with illumination defects on each vehicle perspective image in the vehicle perspective image set and the amount of shooting feature loss;

[0010] Based on the identification feature and the preset identification feature, an identification comparison is performed to determine whether the target vehicle is parked in a standardized manner.

[0011] In some embodiments, performing distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain the vehicle perspective correction image set specifically includes:

[0012] Selecting a vehicle perspective image from the vehicle perspective image set;

[0013] Performing distortion correction on each pixel in the vehicle perspective image to obtain a vehicle perspective adjustment image of the vehicle perspective image;

[0014] Repeat the above steps to obtain vehicle perspective calibration images of the remaining vehicle perspective images in the vehicle perspective image set;

[0015] All vehicle perspective calibration images are combined to obtain a vehicle perspective calibration image set.

[0016] In some embodiments, determining the color sudden change block corresponding to the selected vehicle perspective adjustment image specifically includes:

[0017] Determining all color space feature values ​​of the selected vehicle perspective adjustment image during a color space conversion process;

[0018] Determine all sudden change identification values ​​of the selected vehicle perspective adjustment image when performing image color sudden change identification according to all color space feature values ​​and each grayscale value in the selected vehicle perspective adjustment image;

[0019] A color sudden change block corresponding to the selected vehicle perspective adjustment image is determined based on all the sudden change identification values.

[0020] In some embodiments, determining the color sudden change block corresponding to the selected vehicle perspective adjustment image based on all sudden change identification values ​​specifically includes:

[0021] Determine all color abrupt selected points based on each abrupt identification value;

[0022] All selected color sudden change points are connected to obtain a color sudden change block corresponding to the selected vehicle perspective adjustment image.

[0023] In some embodiments, determining the amount of shooting feature loss when shooting the target vehicle under the influence of lighting according to the image lighting information and the recognition cost specifically includes:

[0024] Acquire an illumination response sequence of each vehicle perspective image in the image illumination information;

[0025] For each vehicle perspective image, performing feature loss analysis on the illumination response sequence corresponding to the vehicle perspective image according to the recognition cost, thereby obtaining all feature loss values ​​corresponding to each vehicle perspective image;

[0026] The shooting feature loss amount when shooting the target vehicle under the influence of lighting is determined based on all feature loss values ​​corresponding to each vehicle perspective image.

[0027] In some embodiments, determining the identification feature of the target vehicle based on all pixels with illumination defects on each vehicle perspective image in the vehicle perspective image set and the shooting feature loss specifically includes:

[0028] Based on the shooting feature loss amount, all illumination defect pixel points with illumination defect on each vehicle perspective image in the vehicle perspective image set are selected, and all illumination defect pixel points of each vehicle perspective image are combined to obtain an illumination defect pixel point set;

[0029] Projecting all illumination defect pixel points in the illumination defect pixel point set to obtain an illumination defect projection surface;

[0030] Determining a calibration quadrilateral of the target vehicle when calibrating the spatial position according to the illumination defect projection surface;

[0031] Performing perspective on the calibration quadrilateral to obtain calibration basis vectors;

[0032] Get the shooting space matrix of the shooting device;

[0033] The identification features of the target vehicle are determined by the shooting space matrix and the calibration basis vectors.

[0034] In some embodiments, the vehicle-perspective image set includes vehicle-perspective images taken at different angles.

[0035] In a second aspect, the present application provides a vehicle identification system for a parking lot, comprising:

[0036] An acquisition module, used to acquire a vehicle perspective image set of a target vehicle in a hospital parking lot;

[0037] A processing module, used for performing distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain a vehicle perspective correction image set;

[0038] The processing module is further used to select a vehicle perspective adjustment image from the vehicle perspective adjustment image set as a selected vehicle perspective adjustment image, determine a color sudden change block corresponding to the selected vehicle perspective adjustment image, extract an identification kernel for color recognition of the selected vehicle perspective adjustment image from the color sudden change block based on a pre-trained convolution model, continue to determine the identification kernels for color recognition of the remaining vehicle perspective adjustment images, and further determine the identification cost of the target vehicle under the image color sudden change through all the identification kernels;

[0039] The processing module is further used to obtain image illumination information of the target vehicle, determine the amount of shooting feature loss when shooting the target vehicle under the influence of illumination according to the image illumination information and the recognition cost, and determine the identification features of the target vehicle based on all pixel points with illumination defects on each vehicle perspective image in the vehicle perspective image set and the amount of shooting feature loss;

[0040] The execution module is used to compare the identification feature with the preset identification feature to determine whether the target vehicle is parked in a standardized manner.

[0041] In a third aspect, the present application provides a computer device, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the computer device executes the above-mentioned parking lot vehicle identification method.

[0042] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions or codes, and when the instructions or codes are executed on a computer, the computer implements the above-mentioned parking lot vehicle identification method when executed.

[0043] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:

[0044] The present application provides a method, system, device and storage medium for identifying a vehicle in a parking lot, which includes obtaining a vehicle perspective image set of a target vehicle in a hospital parking lot; performing distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain a vehicle perspective correction image set; selecting a vehicle perspective correction image from the vehicle perspective correction image set as a selected vehicle perspective correction image, determining a color sudden change block corresponding to the selected vehicle perspective correction image, extracting an identification kernel for color identification of the selected vehicle perspective correction image from the color sudden change block based on a pre-trained convolutional model, and continuing Determine the recognition kernel for color recognition of the remaining vehicle perspective adjustment images, and then determine the recognition cost of the target vehicle under the sudden color change of the image through all the recognition kernels; obtain the image illumination information of the target vehicle, determine the amount of shooting feature loss when shooting the target vehicle under the influence of illumination according to the image illumination information and the recognition cost, and determine the identification feature of the target vehicle based on all pixel points with illumination defects on each vehicle perspective image in the vehicle perspective image set and the amount of shooting feature loss; perform identification comparison based on the identification feature and the preset identification feature to determine whether the target vehicle is parked in a standardized manner.

[0045] It can be seen that in the present application, when marking the target vehicle, it is possible to accurately identify the position of the vehicle outline and the parking space frame under the interference of complex color features, so as to standardize the parking status of the vehicle based on the identification result to avoid the situation where the vehicle is parked randomly; first, the color space conversion is performed on the vehicle perspective calibration image after distortion adjustment to obtain more obvious color change characteristics, and then the color sudden change block is effectively extracted through the result of color space conversion, which improves the color recognition of the parking space frame and the vehicle outline, thereby avoiding the interference caused by the complex color features of the parking space frame and the vehicle outline; secondly, based on the color sudden change block The determined recognition cost can effectively associate the changing features in each vehicle perspective adjustment image, thereby improving the accuracy of the identification of the target vehicle when it is parked in a standardized manner; in addition, the image lighting information is used to determine the identification features of the target vehicle when it is parked, which can effectively identify the parking posture of the target vehicle; finally, based on the identification features, it is judged whether the target vehicle is parked in a standardized manner, which effectively avoids the random parking of vehicles and improves the overall operating efficiency and patient experience of the hospital parking lot; in summary, the technical solution provided in the present application can accurately identify the position of the vehicle contour and the parking space border under the interference of complex color features, so as to standardize the parking status of the vehicle based on the identification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0047] Figure 1 is an exemplary flow chart of a method for identifying a vehicle in a parking lot according to some embodiments of the present application;

[0048] Figure 2 is an exemplary flow chart of determining a color sudden change block according to some embodiments of the present application;

[0049] Figure 3 is an exemplary flow chart of determining identification features according to some embodiments of the present application;

[0050] Figure 4 is a schematic diagram of exemplary hardware and / or software of a vehicle identification system for a parking lot according to some embodiments of the present application;

[0051] Figure 5 It is a structural schematic diagram of a computer device for implementing a method for identifying a vehicle in a parking lot according to some embodiments of the present application. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] The core of the present application is to obtain a vehicle perspective image set of a target vehicle in a hospital parking lot; perform distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain a vehicle perspective correction image set; select a vehicle perspective correction image from the vehicle perspective correction image set as a selected vehicle perspective correction image, determine a color sudden change block corresponding to the selected vehicle perspective correction image, extract a recognition kernel for color recognition of the selected vehicle perspective correction image from the color sudden change block based on a pre-trained convolution model, continue to determine the recognition kernel for color recognition of the remaining vehicle perspective correction images, and then determine the recognition cost of the target vehicle under the sudden color change of the image through all the recognition kernels; obtain image illumination information of the target vehicle, determine the amount of shooting feature loss when shooting the target vehicle under the influence of illumination according to the image illumination information and the recognition cost, determine the identification feature of the target vehicle based on all pixel points with illumination defects on each vehicle perspective image in the vehicle perspective image set and the amount of shooting feature loss; perform identification comparison based on the identification feature and the preset identification feature to determine whether the target vehicle is parked in a standardized manner. The above scheme is based on identification features and can accurately identify the position of the vehicle outline and the parking space border under the interference of complex color features, so as to standardize the parking status of the vehicle based on the identification results.

[0054] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 , which is an exemplary flow chart of a method for identifying a vehicle in a parking lot according to some embodiments of the present application. The method 100 for identifying a vehicle in a parking lot mainly includes the following steps:

[0055] In step 101, a vehicle perspective image set of a target vehicle in a hospital parking lot is obtained.

[0056] In a specific implementation, multiple stereo cameras are used to shoot vehicle perspective images of the target vehicle in the hospital parking lot from different perspectives to obtain a vehicle perspective image set of the target vehicle in the hospital parking lot, wherein the vehicle perspective image set is composed of vehicle perspective images from different shooting angles.

[0057] It should be noted that the vehicle perspective image in the present application represents the image of the target vehicle captured by the stereo camera in the hospital parking lot through different shooting angles. By acquiring the vehicle perspective image, the color characteristics and texture characteristics of the target vehicle can be effectively identified.

[0058] In step 102, distortion correction is performed on each vehicle perspective image in the vehicle perspective image set to obtain a vehicle perspective correction image set.

[0059] In some embodiments, the following method may be used to perform distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain the vehicle perspective correction image set, namely:

[0060] Selecting a vehicle perspective image from the vehicle perspective image set;

[0061] Performing distortion correction on each pixel in the vehicle perspective image to obtain a vehicle perspective adjustment image of the vehicle perspective image;

[0062] Repeat the above steps to obtain vehicle perspective calibration images of the remaining vehicle perspective images in the vehicle perspective image set;

[0063] All vehicle perspective calibration images are combined to obtain a vehicle perspective calibration image set.

[0064] In specific implementation, image processing software in the prior art can be used to perform distortion correction on each pixel in the vehicle perspective image to obtain a vehicle perspective adjustment image of the vehicle perspective image. For example, image processing software Open CV can be used to perform distortion correction on each pixel in the vehicle perspective image to obtain a vehicle perspective adjustment image of the vehicle perspective image, which will not be repeated here. In addition, in other embodiments, other image processing software can also be used to perform distortion correction on each pixel in the vehicle perspective image, such as Photoshop, which is not limited here.

[0065] It should be noted that the vehicle perspective adjustment image set includes multiple vehicle perspective adjustment images. The vehicle perspective adjustment image in this application represents a vehicle perspective image with geometric distortion eliminated after being processed by image processing software. The determination of the vehicle perspective adjustment image can effectively improve the recognition of the target vehicle and highlight the detailed features of the target vehicle, thereby increasing the accuracy and robustness of subsequent operations.

[0066] It should also be noted that the distortion correction in the present application refers to the process of removing the geometric distortion caused by shooting in the vehicle perspective image. The geometric distortion refers to the distortion phenomenon in which the straight lines and shapes in the image are bent and deformed due to the optical characteristics of the camera lens or the shooting angle. Common geometric distortions include radial distortion (such as barrel distortion and pincushion distortion) and perspective distortion. The distortion correction in this embodiment is a method of distortion correction. In other embodiments, other correction methods can also be used to perform distortion correction on each vehicle perspective image in the vehicle perspective image set, which is not limited here.

[0067] In step 103, a vehicle perspective adjustment image is selected from the vehicle perspective adjustment image set as the selected vehicle perspective adjustment image, and a color sudden change block corresponding to the selected vehicle perspective adjustment image is determined. Based on a pre-trained convolutional model, an identification kernel for color recognition of the selected vehicle perspective adjustment image is extracted from the color sudden change block, and identification kernels for color recognition of the remaining vehicle perspective adjustment images are further determined, and then the identification cost of the target vehicle under the sudden color change of the image is determined through all the identification kernels.

[0068] Preferably, in some embodiments, reference Figure 2 As shown in FIG. 1 , this figure is an exemplary flow chart of determining a color sudden change block in some embodiments of the present application. In this embodiment, determining a color sudden change block corresponding to the selected vehicle perspective adjustment image can be implemented by the following steps:

[0069] First, in step 1031, all color space feature values ​​of the selected vehicle perspective adjustment image during the color space conversion process are determined;

[0070] Then, in step 1032, all sudden change identification values ​​of the selected vehicle perspective adjustment image when performing image color sudden change identification are determined according to all color space feature values ​​and each grayscale value in the selected vehicle perspective adjustment image;

[0071] Finally, in step 1033 , the color sudden change block corresponding to the selected vehicle perspective adjustment image is determined based on all the sudden change identification values.

[0072] In a specific implementation, all color space feature values ​​of the selected vehicle perspective adjustment image during the color space conversion process are determined, that is, multiple channel pixel values ​​of each pixel point are extracted from the selected vehicle perspective adjustment image, and the multiple channel pixel values ​​include pixel values ​​of the three color channels of red, green and blue in the image. For each pixel point, the average of the corresponding multiple channel pixel values ​​is used as the color space feature value of the pixel point, thereby obtaining all the color space feature values.

[0073] It should be noted that the color space conversion in this embodiment refers to the process of converting the selected vehicle perspective adjustment image from one color space to another color space, which will not be described in detail here.

[0074] It should also be noted that, in this embodiment, multiple channel pixel values ​​represent pixel values ​​of each color channel, and the color channels are the red, green, and blue channels in the image. In addition, the color space feature values ​​in this application represent the color features of multiple channel pixel values ​​of the selected vehicle perspective adjustment image in the new color space. The color space feature values ​​can effectively characterize the color space features of the selected vehicle perspective adjustment image. Therefore, by determining the color space feature values, the accuracy of identifying subsequent image color sudden change areas can be improved.

[0075] It should be noted that in the present application, one pixel point in the selected vehicle perspective adjustment image corresponds to a color space characteristic value, one pixel point corresponds to a grayscale value, and one pixel point corresponds to pixel values ​​of three color channels: red, green, and blue.

[0076] In a specific implementation, all sudden change identification values ​​of the selected vehicle perspective adjustment image when identifying sudden color changes in the image are determined based on all color space feature values ​​and each grayscale value in the selected vehicle perspective adjustment image, that is: the grayscale value and the color space feature value corresponding to each pixel point in the selected vehicle perspective adjustment image are difference calculated, and the results of each difference calculation are used as all sudden change identification values ​​of the selected vehicle perspective adjustment image when identifying sudden color changes in the image.

[0077] It should be noted that the sudden change identification value in the present application represents a parameter used to identify a sudden change in image color in the perspective adjustment image of the selected vehicle. The sudden change identification value can be used to effectively identify the area where the image color in the perspective adjustment image of the selected vehicle changes suddenly, thereby improving the accuracy of target vehicle identification.

[0078] It should also be noted that the sudden color change in the present application refers to the phenomenon that an obvious color change occurs in the image adjusted by the selected vehicle viewing angle.

[0079] In addition, it should be noted that in the present application, each pixel point in the vehicle perspective adjustment image corresponds to a color space feature value, and each pixel point also corresponds to a sudden change identification value.

[0080] Preferably, in some embodiments, the color sudden change block corresponding to the selected vehicle perspective adjustment image may be determined based on all sudden change identification values ​​in the following manner, namely:

[0081] Determine all color abrupt selected points based on each abrupt identification value;

[0082] All selected color sudden change points are connected to obtain a color sudden change block corresponding to the selected vehicle perspective adjustment image.

[0083] In the specific implementation, all the color sudden change selected points are determined based on each sudden change identification value, that is: for each sudden change identification value, when the sudden change identification value is a non-negative value, the pixel point corresponding to the sudden change identification value is identified as the color sudden change selected point, otherwise the pixel point corresponding to the sudden change identification value is not processed, thereby obtaining all the color sudden change selected points.

[0084] It should be noted that, in this embodiment, the selected point of sudden color change represents the pixel point selected as the place where the color obviously changes in the selected vehicle perspective adjustment image. By selecting the pixel point where the color obviously changes in the selected vehicle perspective adjustment image, the sudden color change area where the color suddenly changes in the selected vehicle perspective adjustment image can be effectively identified.

[0085] In a specific implementation, all the selected color change points are connected to obtain the color change block corresponding to the selected vehicle perspective adjustment image, that is, all the selected color change points are connected by the region growing algorithm in the prior art to obtain the color change block corresponding to the selected vehicle perspective adjustment image. For example, starting from one color change selected point, the remaining color change selected points are connected in a recursive form to obtain the color change block corresponding to the selected vehicle perspective adjustment image. The recursion is a process of performing operations in a preset order.

[0086] It should be noted that the region growing algorithm in this embodiment is an existing technology for image segmentation, which will not be described in detail here.

[0087] It should also be noted that the color sudden change block in the present application is used to identify the area with obvious visual color changes in the perspective adjustment image of the selected vehicle. In the vehicle identification technology in the hospital parking lot, determining the color sudden change block can extract the characteristic information of the perspective adjustment image of the selected vehicle, so as to more accurately compare and identify the current parking state of the target vehicle with the standard parking state during standard parking. The color sudden change block includes multiple color space feature values, that is, the color space feature value corresponding to each pixel point. In addition, in other embodiments, other implementation methods can also be used to determine the color sudden change block where the color of each vehicle perspective adjustment image in the vehicle perspective adjustment image set changes suddenly, which is not limited here.

[0088] In some embodiments, the recognition kernel for color recognition of the selected vehicle perspective adjustment image extracted from the color sudden change block based on the pre-trained convolution model may be used in the following manner, namely:

[0089] Call the pre-trained convolutional model;

[0090] The convolution model is used to extract the recognition kernel of the selected vehicle perspective adjustment image from each color sudden change block for color recognition.

[0091] In the specific implementation, the pre-trained convolution model is directly called by the computer. In the present application, a convolutional neural network model is used as the pre-trained convolution model. For example, the Visual Geometry Group (VGG) model can be used as the pre-trained convolution model. In addition, other classic models widely used in image processing tasks or other customized convolutional neural networks can be used as pre-trained convolution models. There is no limitation here. It should be noted that the pre-trained convolution model in this embodiment is obtained by training with a large amount of historical image data as training samples, which will not be repeated here.

[0092] It should also be noted that the identification kernel in the present application represents a matrix that performs the same spatial mapping of all color space feature values ​​within the color sudden change block. The matrix is ​​used to comprehensively measure the feature distribution in the selected vehicle perspective adjustment image so as to be used to identify and estimate the target vehicle under the sudden color change of the image. In addition, in other embodiments, other methods can be used to obtain the identification kernel, which will not be repeated here.

[0093] It should be noted that in the present application, one vehicle perspective adjustment image corresponds to one recognition kernel, and the vehicle perspective adjustment image is shot in accordance with a prescribed shooting order, so adjacent vehicle perspective adjustment images correspond to adjacent recognition kernels.

[0094] In some embodiments, the recognition cost of the target vehicle under the sudden color change of the image can be determined by all recognition kernels in the following manner, namely:

[0095] Perform correlation calculation on all adjacent identification kernels to obtain all sudden change correlation identification values;

[0096] The sum of all sudden change associated recognition values ​​is taken as the recognition cost of the target vehicle under sudden color change of the image.

[0097] It should be noted that in other embodiments, other calculation methods may be used to calculate the recognition cost of the target vehicle under the sudden color change of the image, which is not limited here.

[0098] It should be noted that, in this embodiment, the sudden change correlation identification value represents a value obtained by performing correlation calculation on different identification kernels.

[0099] In specific implementation, the Euclidean distance may be used to perform association calculation on all adjacent identification kernels. In other embodiments, other association calculation methods may be used to perform similarity calculation, which is not limited here.

[0100] It should be noted that the recognition cost in the present application represents an amount used to measure the degree of correlation loss of color change blocks between all vehicle perspective adjustment images of the target vehicle, that is, the greater the recognition cost, the greater the degree of correlation loss of color change blocks between all vehicle perspective adjustment images of the target vehicle, and the smaller the recognition cost, the smaller the degree of correlation loss of color change blocks between all vehicle perspective adjustment images of the target vehicle. By determining the recognition cost, the changing features in each vehicle perspective adjustment image are effectively associated, thereby improving the accuracy of identifying whether the target vehicle is parked in a standardized manner.

[0101] In step 104, image illumination information of the target vehicle is obtained, and the amount of shooting feature loss when shooting the target vehicle under the influence of illumination is determined based on the image illumination information and the recognition cost, and the identification features of the target vehicle are determined based on all pixel points with illumination defects on each vehicle perspective image in the vehicle perspective image set and the amount of shooting feature loss.

[0102] In some embodiments, the image illumination information of the target vehicle may be obtained in the following manner, namely:

[0103] Selecting a vehicle perspective image from the vehicle perspective image set, analyzing the influence of illumination on each pixel in the vehicle perspective image, and obtaining an illumination response sequence of the vehicle perspective image;

[0104] Repeat the above steps to obtain a lighting response sequence of the remaining vehicle perspective images in the vehicle perspective image set;

[0105] The illumination response sequences in each vehicle perspective image are combined to obtain the image illumination information when shooting the target vehicle in the hospital parking lot.

[0106] It should be noted that, in the process of shooting an image, certain lighting conditions usually need to be met. Images obtained under different lighting conditions will show different image brightness. Therefore, the grayscale value in the image can be used to characterize the light response of the pixel. As a preferred embodiment, each pixel in the vehicle perspective image is analyzed for the influence of lighting to obtain the light response sequence of the vehicle perspective image, that is, the vehicle perspective image is input into the image processing software to extract the grayscale value of each pixel, and the grayscale value of each pixel is used as the light response, and the grayscale value of each pixel is combined to obtain the light response sequence of the vehicle perspective image. Specifically, the vehicle perspective image can be input into Open CV to extract the grayscale of each pixel, which will not be repeated here.

[0107] It should be noted that the image illumination information in the present application includes an illumination response sequence of multiple vehicle perspective images, and the illumination response sequence includes multiple illumination response quantities. In the present embodiment, the illumination response quantity represents the response parameter of the pixel point to the light intensity under different illumination conditions. Specifically, this can be expressed as the grayscale value corresponding to the pixel point under specific illumination. By determining the illumination response quantity, the characteristics of the image under the influence of illumination can be more robustly characterized, so that subsequent operations and calculations are easier to implement.

[0108] Preferably, in some embodiments, the following method may be used to determine the amount of shooting feature loss when shooting the target vehicle under the influence of lighting according to the image lighting information and the recognition cost, namely:

[0109] Acquire an illumination response sequence of each vehicle perspective image in the image illumination information;

[0110] For each vehicle perspective image, performing feature loss analysis on the illumination response sequence corresponding to the vehicle perspective image according to the recognition cost, thereby obtaining all feature loss values ​​corresponding to each vehicle perspective image;

[0111] The shooting feature loss amount when shooting the target vehicle under the influence of lighting is determined based on all feature loss values ​​corresponding to each vehicle perspective image.

[0112] In some embodiments, for each vehicle perspective image, a feature loss analysis is performed on the illumination response sequence corresponding to the vehicle perspective image according to the recognition cost, and then all feature loss values ​​corresponding to each vehicle perspective image are obtained in the following manner, namely:

[0113] Select a vehicle perspective image, and sum all illumination response quantities in the illumination response sequence corresponding to the vehicle perspective image with the recognition cost to obtain all feature loss values ​​corresponding to the vehicle perspective image;

[0114] Repeat the above steps to obtain all feature loss values ​​corresponding to the remaining vehicle perspective images.

[0115] It should be noted that in other embodiments, other calculation methods may be used to calculate the characteristic loss value, which is not limited here.

[0116] It should be noted that the feature loss value in this embodiment represents the light loss value recorded after capturing the light intensity at the pixel point in the image, where each pixel point corresponds to a feature loss value. Determining the feature loss value helps to optimize the brightness and contrast of the image and improve the robustness of image processing under the influence of light.

[0117] In specific implementation, the shooting feature loss amount when shooting the target vehicle under the influence of lighting is determined based on all feature loss values ​​corresponding to each vehicle perspective image, that is: all feature loss values ​​corresponding to each vehicle perspective image are aggregated to obtain the shooting feature loss amount when shooting the target vehicle under the influence of lighting. Specifically, the mean method can be used to aggregate all feature loss values ​​corresponding to each vehicle perspective image to obtain the shooting feature loss amount when shooting the target vehicle under the influence of lighting. It will not be repeated here. In addition, in other embodiments, other aggregation methods can be used to aggregate all feature loss values ​​corresponding to each vehicle perspective image, which are not limited here.

[0118] It should be noted that the shooting feature loss amount in the present application represents the feature loss of the captured vehicle perspective image caused by lighting changes. This value can be used to measure the changes in lighting effects between images from different vehicle perspectives during shooting. By determining the shooting feature loss amount, the stability of the image feature information can be ensured.

[0119] Preferably, in some embodiments, reference Figure 3 As shown in FIG. 1 , this figure is an exemplary flow chart for determining identification features in some embodiments of the present application. In this embodiment, the identification features of the target vehicle are determined based on all pixels with illumination defects on each vehicle perspective image in the vehicle perspective image set and the shooting feature loss amount, which can be implemented by the following steps:

[0120] First, in step 1041, all illumination defect pixel points with illumination defect on each vehicle perspective image in the vehicle perspective image set are selected based on the shooting feature loss amount, and all illumination defect pixel points of each vehicle perspective image are combined to obtain an illumination defect pixel point set;

[0121] Then, in step 1042, all illumination defect pixel points in the illumination defect pixel point set are projected to obtain an illumination defect projection surface;

[0122] Secondly, in step 1043, a calibration quadrilateral of the target vehicle when calibrating the spatial position is determined according to the illumination defect projection surface;

[0123] In addition, in step 1044, the calibration quadrilateral is perspectively analyzed to obtain calibration basis vectors;

[0124] Also, in step 1045, a shooting space matrix of the shooting device is obtained;

[0125] Finally, in step 1046, the identification features of the target vehicle are determined by using the shooting space matrix and the calibration basis vectors.

[0126] It should be noted that in this embodiment, the grayscale value corresponding to each pixel in the vehicle perspective image is used as the illumination response, and each illumination response corresponds to a feature loss value. Therefore, a pixel also corresponds to a feature loss value, which will not be repeated here.

[0127] In some embodiments, all illumination defect pixels with illumination defect on each vehicle perspective image in the vehicle perspective image set are selected based on the shooting feature loss amount in the following manner, namely:

[0128] Acquire the shooting feature loss amount and each vehicle perspective image in the vehicle perspective image set;

[0129] Selecting a vehicle perspective image from the vehicle perspective image set;

[0130] Extracting a feature loss value corresponding to each pixel in the vehicle perspective image, comparing the feature loss value corresponding to each pixel with the shooting feature loss amount, treating the pixel points corresponding to each feature loss value greater than the shooting feature loss amount as illumination defect pixel points with illumination defect on the vehicle perspective image, and not processing the pixel points corresponding to each feature loss value less than or equal to the shooting feature loss amount;

[0131] Repeat the above steps to obtain all illumination defect pixel points with illumination defect on the remaining vehicle perspective images in the vehicle perspective image set.

[0132] It should be noted that, in this embodiment, the illumination defect pixel points refer to the pixel points in the vehicle perspective image that have a poor effect of capturing the illumination intensity due to illumination changes.

[0133] In a specific implementation, all illumination defect pixel points in the illumination defect pixel point set are projected to obtain an illumination defect projection surface, that is, the spatial coordinates corresponding to each illumination defect pixel point in the illumination defect pixel point set are extracted, the spatial coordinates of each illumination defect pixel point are two-dimensionally projected to obtain all illumination projection points, all illumination projection points are geometrically connected, and then the illumination defect projection surface is obtained.

[0134] In specific implementation, the spatial coordinates corresponding to each illumination defect pixel point in the illumination defect pixel point set can be directly extracted by a computer, which will not be described in detail here.

[0135] It should be noted that, in this embodiment, two-dimensional projection is the process of mapping the spatial coordinates of all illumination-deficient pixel points onto a two-dimensional plane. The purpose of performing two-dimensional projection on the spatial coordinates of each illumination-deficient pixel point is to further perform geometric analysis. The illumination projection point represents the projection form of the spatial coordinates of the illumination-deficient pixel point on the two-dimensional plane.

[0136] It should also be noted that the illumination defect projection surface in the present application represents a two-dimensional plane area obtained by connecting all illumination projection points through a geometric connection method. In specific implementation, the convex hull connection method in the prior art can be used to connect all illumination projection points. In addition, in other embodiments, other geometric connection methods can also be used to connect all illumination projection points.

[0137] In specific implementation, the calibration quadrilateral of the target vehicle when calibrating the spatial position is determined according to the illumination defect projection surface, that is: the center point of the illumination defect projection surface is obtained, the distance between the center point and each illumination projection point in the illumination defect projection surface is calculated to obtain all projection distance values, all projection distance values ​​are arranged in ascending order, and the illumination projection points corresponding to the last four projection distance values ​​after arrangement are used as four geometric base points, and the four geometric base points are connected to obtain a calibration quadrilateral.

[0138] It should be noted that the projection distance value in this embodiment represents the distance between the center point of the illumination defect projection surface and the illumination projection point in the illumination defect projection surface. The geometric base point in this embodiment represents the reference point used as the geometric connection. The calibrated quadrilateral in this embodiment is a two-dimensional plane quadrilateral determined by connecting the geometric base points.

[0139] In specific implementation, the calibration quadrilateral can be perspectived by the perspective principle to obtain calibration basis vectors. For example, the calibration quadrilateral can be perspectived by the vanishing point perspective method in the prior art to obtain calibration basis vectors, which will not be described in detail here.

[0140] It should be noted that the calibration basis vectors in this embodiment are used to characterize the spatial relationship between the target vehicle position and the shooting device in perspective projection. By determining the calibration basis vectors, the parking area and the vehicle body contour can be identified more accurately, thereby identifying whether the subsequent parking status of the target vehicle is standardized.

[0141] In specific implementation, the shooting space matrix of the shooting device can be obtained through the vehicle identification database of the hospital parking lot. The shooting space matrix is ​​a matrix that describes the spatial position and shooting direction of the shooting device. The shooting space matrix is ​​specifically determined by the position of the shooting device, which will not be repeated here. In the standardized identification of the parking posture of the target vehicle in the parking lot, the determination of the shooting space matrix can effectively identify the spatial position relationship between the current target vehicle and the shooting device, thereby obtaining accurate vehicle position and posture data, so as to improve the accuracy and reliability of subsequent standardized identification of the target vehicle.

[0142] It should be noted that the identification feature in the present application represents a normative measurement of the parking posture of the target vehicle in the parking lot. By determining the normative identification feature, the parking posture of the target vehicle can be effectively identified in a normative manner, thereby improving the accuracy and reliability of the parking normative identification of the target vehicle. As a preferred embodiment, the identification feature of the target vehicle is determined by the shooting space matrix and the calibration basis vector. The following method is adopted, namely: first, the calibration basis vector is transposed, and then the transposed calibration basis vector is multiplied by the shooting space matrix, and finally the calculated vector is used as the identification feature of the target vehicle when parking. In addition, in other embodiments, other calculation methods can be used to calculate the identification feature, which is not limited here.

[0143] In step 105, a comparison is performed based on the identification feature and a preset identification feature to determine whether the target vehicle is parked in a standardized manner.

[0144] In some embodiments, based on the identification feature and the preset identification feature, the following method can be used to determine whether the target vehicle is parked in a standardized manner, namely:

[0145] Determining the cosine similarity between the identification feature and a preset identification feature;

[0146] The cosine similarity is compared with a preset similarity threshold to determine whether the target vehicle is parked in a standardized manner.

[0147] It should be noted that, in this embodiment, cosine similarity is an existing technology commonly used for identification comparison between features. By determining the cosine similarity of two features, the spatial similarity between the two features can be quantified. The value of cosine similarity is between -1 and 1.

[0148] In specific implementation, the cosine similarity is compared with a preset similarity threshold to determine whether the target vehicle is parked in a standardized manner, that is, the cosine similarity is compared with the preset similarity threshold. When the cosine similarity is greater than the preset similarity threshold, it indicates that the target vehicle is in a standardized parking state at this time, and no subsequent operation is performed; when the cosine similarity is less than or equal to the preset similarity threshold, it indicates that the target vehicle is not parked in a standardized manner at this time, and is determined to be an illegally parked vehicle, and the vehicle identification system is triggered to mark the target vehicle as illegally parked, so as to send a notification to relevant management personnel and take corresponding measures.

[0149] It should be noted that the preset identification feature in the present application is a pre-set standard identification feature, which can be set according to the actual needs of the hospital parking lot and will not be elaborated here.

[0150] It should also be noted that, in this embodiment, the preset similarity threshold is a pre-set standard cosine similarity, which can be used to define the similarity effect between two features. The preset similarity threshold is 1, indicating that the two features are completely similar, that is, when the preset similarity threshold is closer to 1, the similarity between the two features is higher. The preset similarity threshold is -1, indicating that the two features are completely dissimilar, that is, when the preset similarity threshold is closer to -

[0151] 1 indicates that the lower the similarity between the two features, specifically, in this embodiment, the preset similarity threshold value is 0.8. In addition, in other embodiments, the preset similarity threshold value can also be taken as other values, which are not limited here.

[0152] It can be seen that in the present application, when marking the target vehicle, it is possible to accurately identify the position of the vehicle outline and the parking space frame under the interference of complex color features, so as to standardize the parking status of the vehicle based on the identification result to avoid the situation where the vehicle is parked randomly; first, the color space conversion is performed on the vehicle perspective calibration image after distortion adjustment to obtain more obvious color change characteristics, and then the color sudden change block is effectively extracted through the result of color space conversion, which improves the color recognition of the parking space frame and the vehicle outline, thereby avoiding the interference caused by the complex color features of the parking space frame and the vehicle outline; secondly, based on the color sudden change block The determined recognition cost can effectively associate the changing features in each vehicle perspective adjustment image, thereby improving the accuracy of the identification of the target vehicle when it is parked in a standardized manner; in addition, the image lighting information is used to determine the identification features of the target vehicle when it is parked, which can effectively identify the parking posture of the target vehicle; finally, based on the identification features, it is judged whether the target vehicle is parked in a standardized manner, which effectively avoids the random parking of vehicles and improves the overall operating efficiency and patient experience of the hospital parking lot; in summary, the technical solution provided in the present application can accurately identify the position of the vehicle contour and the parking space border under the interference of complex color features, so as to standardize the parking status of the vehicle based on the identification results.

[0153] In addition, in another aspect of the present application, in some embodiments, the present application provides a vehicle identification system for a parking lot, referring to Figure 4 , which is a schematic diagram of exemplary hardware and / or software of a parking lot vehicle identification system according to some embodiments of the present application, the parking lot vehicle identification system 400 includes: an acquisition module 401, a processing module 402 and an execution module 403, which are described as follows:

[0154] Acquisition module 401, in this application, acquisition module 401 is mainly used to acquire a vehicle perspective image set of a target vehicle in a hospital parking lot;

[0155] The processing module 402 in the present application is mainly used to perform distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain a vehicle perspective correction image set;

[0156] It should be noted that the processing module 402 in the present application is also used to select a vehicle perspective adjustment image from the vehicle perspective adjustment image set as the selected vehicle perspective adjustment image, determine the color sudden change block corresponding to the selected vehicle perspective adjustment image, extract the recognition kernel for color recognition of the selected vehicle perspective adjustment image from the color sudden change block based on the pre-trained convolution model, continue to determine the recognition kernel for color recognition of the remaining vehicle perspective adjustment images, and then determine the recognition cost of the target vehicle under the image color sudden change through all the recognition kernels;

[0157] In addition, the processing module 402 in the present application is also used to obtain image illumination information of the target vehicle, determine the amount of shooting feature loss when shooting the target vehicle under the influence of illumination according to the image illumination information and the recognition cost, and determine the identification features of the target vehicle based on all pixel points with illumination defects on each vehicle perspective image in the vehicle perspective image set and the amount of shooting feature loss;

[0158] Execution module 403, in the present application, execution module 403 is mainly used to compare the identification features with the preset identification features to determine whether the target vehicle is parked in a standardized manner.

[0159] The above describes in detail the examples of the method, system, device and storage medium for vehicle identification in the parking lot provided by the embodiments of the present application. It can be understood that in order to realize the above functions, the corresponding device includes a hardware structure and / or software module corresponding to the execution of each function. It should be easily appreciated by those skilled in the art that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present application.

[0160] In some embodiments, the present application also provides a computer device, which includes a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the computer device executes the above-mentioned parking lot vehicle identification method.

[0161] In some embodiments, reference Figure 5, the dotted line in the figure indicates that the unit or the module is optional, and the figure is a schematic diagram of the structure of a computer device of a parking lot vehicle identification method provided in an embodiment of the present application. The parking lot vehicle identification method described in the above embodiment can be Figure 5 The computer device 500 shown in the figure is implemented, and the computer device 500 includes at least one processor 501, a memory 502 and at least one communication unit 505. The computer device 500 can be a terminal device, a server or a chip.

[0162] The processor 501 may be a general-purpose processor or a special-purpose processor. For example, the processor 501 may be a central processing unit (CPU), which may be used to control the computer device 500, execute software programs, and process data of the software programs. The computer device 500 may also include a communication unit 505 to implement signal input (reception) and output (transmission).

[0163] For example, the computer device 500 may be a chip, the communication unit 505 may be an input and / or output circuit of the chip, or the communication unit 505 may be a communication interface of the chip, and the chip may be a component of a terminal device, a network device, or other devices.

[0164] For another example, the computer device 500 may be a terminal device or a server, and the communication unit 505 may be a transceiver of the terminal device or the server, or the communication unit 505 may be a transceiver circuit of the terminal device or the server.

[0165] The computer device 500 may include one or more memories 502, on which a program 504 is stored. The program 504 can be executed by the processor 501 to generate instructions 503, so that the processor 501 performs the method described in the above method embodiment according to the instructions 503. Optionally, data (such as a target audit model) can also be stored in the memory 502. Optionally, the processor 501 can also read the data stored in the memory 502, and the data can be stored at the same storage address as the program 504, or the data can be stored at a different storage address from the program 504.

[0166] The processor 501 and the memory 502 may be provided separately or integrated together, for example, integrated on a system on chip (SOC) of the terminal device.

[0167] It should be understood that each step of the above method embodiment can be completed by a hardware-based logic circuit or software-based instructions in the processor 501. The processor 501 can be a central processing unit, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, such as discrete gates, transistor logic devices or discrete hardware components.

[0168] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0169] For example, in some embodiments, the present application also provides a computer-readable storage medium, in which instructions or codes are stored. When the instructions or codes are executed on a computer, the computer implements the above-mentioned parking lot vehicle identification method when executing.

[0170] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0171] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for identifying vehicles in a parking lot, characterized in that: The steps include: Obtain a vehicle-view image set of a target vehicle in a hospital parking lot; Performing distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain a vehicle perspective correction image set; A vehicle perspective adjustment image is selected from the vehicle perspective adjustment image set as a selected vehicle perspective adjustment image, a color sudden change block corresponding to the selected vehicle perspective adjustment image is determined, and an identification kernel for color recognition of the selected vehicle perspective adjustment image is extracted from the color sudden change block based on a pre-trained convolution model, wherein the identification kernel represents a matrix that performs the same spatial mapping on all color space feature values ​​in the color sudden change block, and the identification kernels for color recognition of the remaining vehicle perspective adjustment images are continuously determined, and then the identification cost of the target vehicle under the image color sudden change is determined through all the identification kernels; Acquire image illumination information of the target vehicle, determine the amount of shooting feature loss when shooting the target vehicle under the influence of illumination according to the image illumination information and the recognition cost, and determine the identification features of the target vehicle based on all pixel points with illumination defects on each vehicle perspective image in the vehicle perspective image set and the amount of shooting feature loss; Based on the identification feature and the preset identification feature, a comparison is performed to determine whether the target vehicle is parked in a standardized manner; Among them, the recognition cost of the target vehicle under the sudden color change of the image is determined by all the recognition kernels in the following way: Perform correlation calculation on all adjacent identification kernels to obtain all sudden change correlation identification values; The sum of all sudden change associated recognition values ​​is used as the recognition cost of the target vehicle under sudden color change of the image; Wherein, determining the shooting feature loss amount when shooting the target vehicle under the influence of lighting according to the image lighting information and the recognition cost specifically includes: Acquire an illumination response sequence of each vehicle perspective image in the image illumination information; For each vehicle perspective image, performing feature loss analysis on the illumination response sequence corresponding to the vehicle perspective image according to the recognition cost, thereby obtaining all feature loss values ​​corresponding to each vehicle perspective image; Determine the shooting feature loss amount when shooting the target vehicle under the influence of lighting based on all feature loss values ​​corresponding to each vehicle perspective image; Wherein, determining the identification features of the target vehicle based on all pixel points with illumination defects on each vehicle perspective image in the vehicle perspective image set and the shooting feature loss amount specifically includes: Based on the shooting feature loss amount, all illumination defect pixel points with illumination defect on each vehicle perspective image in the vehicle perspective image set are selected, and all illumination defect pixel points of each vehicle perspective image are combined to obtain an illumination defect pixel point set; Projecting all illumination defect pixel points in the illumination defect pixel point set to obtain an illumination defect projection surface; Determining a calibration quadrilateral of the target vehicle when calibrating the spatial position according to the illumination defect projection surface; Performing perspective on the calibration quadrilateral to obtain calibration basis vectors; Get the shooting space matrix of the shooting device; The identification features of the target vehicle are determined by the shooting space matrix and the calibration basis vectors.

2. The method according to claim 1, characterized in that Performing distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain the vehicle perspective correction image set specifically includes: Selecting a vehicle perspective image from the vehicle perspective image set; Performing distortion correction on each pixel in the vehicle perspective image to obtain a vehicle perspective adjustment image of the vehicle perspective image; Repeat the above steps to obtain vehicle perspective calibration images of the remaining vehicle perspective images in the vehicle perspective image set; All vehicle perspective calibration images are combined to obtain a vehicle perspective calibration image set.

3. The method according to claim 1, characterized in that Determining the color sudden change block corresponding to the selected vehicle perspective adjustment image specifically includes: Determining all color space feature values ​​of the selected vehicle perspective adjustment image during a color space conversion process; Determine all sudden change identification values ​​of the selected vehicle perspective adjustment image when performing image color sudden change identification according to all color space feature values ​​and each grayscale value in the selected vehicle perspective adjustment image; A color sudden change block corresponding to the selected vehicle perspective adjustment image is determined based on all the sudden change identification values.

4. The method according to claim 3, characterized in that Determining the color sudden change block corresponding to the selected vehicle perspective adjustment image based on all sudden change identification values ​​specifically includes: Determine all color abrupt selected points based on each abrupt identification value; All selected color sudden change points are connected to obtain a color sudden change block corresponding to the selected vehicle perspective adjustment image.

5. The method according to claim 1, characterized in that The vehicle perspective image set includes vehicle perspective images taken at different shooting angles.

6. A vehicle identification system for a parking lot, which uses the method according to any one of claims 1 to 5 to identify vehicles, characterized in that: The parking lot's vehicle identification system includes: An acquisition module, used to acquire a vehicle perspective image set of a target vehicle in a hospital parking lot; A processing module, used for performing distortion correction on each vehicle perspective image in the vehicle perspective image set to obtain a vehicle perspective correction image set; The processing module is further used to select a vehicle perspective adjustment image from the vehicle perspective adjustment image set as a selected vehicle perspective adjustment image, determine a color sudden change block corresponding to the selected vehicle perspective adjustment image, extract an identification kernel for color recognition of the selected vehicle perspective adjustment image from the color sudden change block based on a pre-trained convolution model, continue to determine the identification kernels for color recognition of the remaining vehicle perspective adjustment images, and further determine the identification cost of the target vehicle under the image color sudden change through all the identification kernels; The processing module is further used to obtain image illumination information of the target vehicle, determine the amount of shooting feature loss when shooting the target vehicle under the influence of illumination according to the image illumination information and the recognition cost, and determine the identification features of the target vehicle based on all pixel points with illumination defects on each vehicle perspective image in the vehicle perspective image set and the amount of shooting feature loss; The execution module is used to compare the identification feature with the preset identification feature to determine whether the target vehicle is parked in a standardized manner.

7. A computer device, characterized in that: The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the computer device executes the vehicle identification method for a parking lot according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions or codes, and when the instructions or codes are executed on a computer, the computer implements the method for identifying vehicles in a parking lot as claimed in any one of claims 1 to 5.

Images