Irregular road parking identification system and equipment

Abstract

The invention discloses a road non-standard parking identification system and equipment, and the system is based on the following identification method: collecting and processing an image of a parking area, and recarving an approximate part of a parking space line; for parking space line extraction, miscellaneous point removal and morphological reconstruction, a parking space line contour line is obtained; calculating an approaching part of the ground and the parking space contour line, and extracting pixel point coordinates to obtain an adjusted parking space contour line; searching and marking a license plate position in the image; carrying out clustering analysis on the contour line, calculating the approximate direction of each part of the contour line, and planning the position of the parking space; and calculating an intersection point of the normal vector equation from the license plate to the road surface and the parking space direction equation, and judging whether the solution of the intersection point is within a value range so as to judge the relative position of the vehicle and the parking space. According to the method, computer vision and an image processing algorithm are utilized, the recognition rate of the vehicle parking behavior and the robustness of the algorithm are effectively improved, the calculation amount is reduced, the manual inspection cost is reduced, and the method can be applied to the field of intelligent traffic.

Description

[0001] This application is a divisional application of the Chinese invention patent application with application number "202410539191.2", application date "2024.04.30", and invention name "Method for identifying irregular parking on roads". Technical Field

[0002] The present invention relates to the field of intelligent transportation technology, in particular to a system and equipment for identifying irregular parking on a road, and in particular to a method, system, equipment and medium for identifying irregular parking on a road. Background Art

[0003] In recent years, the number of motor vehicles in my country has continued to grow, and cars have become an indispensable means of transportation for people's daily travel. The problem of parking difficulties has become increasingly prominent. Especially in busy commercial areas, residential areas and near public transportation hubs, finding a suitable parking space has become a headache for many drivers. The planning of roadside parking spaces can greatly alleviate the pressure of parking and has been increasingly promoted. At the current stage, the supervision of roadside parking spaces in most cities in my country is still mainly manual, which consumes a lot of labor costs, is extremely inefficient, and it is difficult to maximize the utilization of parking space resources. Especially during peak parking periods, errors are often prone to occur. Therefore, the development of an efficient parking space and image recognition technology is imminent. However, parking space and vehicle image recognition technology still faces some challenges in practical applications. For example, factors such as different vehicle models, lighting conditions, and obstructions may affect the accuracy of image recognition. In addition, with the continuous development of technology and the increasing diversification of user needs, how to further improve recognition accuracy, optimize user experience, and reduce system costs have also become hot spots and difficulties in research in this field.

[0004] At present, there are two main methods for managing roadside parking spaces: geomagnetism-based and video-based. The parking space detection method based on geomagnetism has high detection accuracy, but it requires the manual installation of geomagnetic detectors, which has high manufacturing costs, and has problems such as difficulty in endurance and susceptibility to signal interference. The parking space detection method based on video does not require additional equipment to be installed, but video detection is greatly affected by the environment, and the accuracy of parking space detection is easily affected at long distances. Therefore, based on this problem, those skilled in the art have proposed a method for identifying irregular parking on roads, which uses computer vision and image processing algorithms to effectively reduce costs, and considers various influencing factors in the actual state of roadside parking spaces, effectively reducing the computational complexity of determining the parking position analysis of roadside vehicles, and greatly reducing the demand for computing resources. Summary of the invention

[0005] Technical issues In order to solve the above problems, the purpose of the present invention is to provide a road irregular parking recognition system and equipment, which uses image processing algorithms based on computer vision to detect parking behaviors, so as to solve the problem of roadside parking space detection. The accuracy of image recognition of the method of the present invention is not affected by factors such as vehicle type, lighting conditions, and obstructions.

[0006] Technical Solution In order to achieve the above object, the present invention provides a method for identifying irregular parking on a road, comprising the following steps: 1) Read the image and convert the RGB image to HSV space; 2) Process the HSV space and reproduce the approximate position of the parking space line; 3) For parking space line extraction, remove noise, perform morphological reconstruction, and obtain the parking space line contour; 4) Calculate the close parts of the ground contour line and the parking space line contour line, extract the pixel point coordinates, and obtain the parking space line contour line adjusted by the ground contour line; 5) Search the license plate position in the image and mark it in the parking space line contour image of step 4); 6) Perform cluster analysis on the contour lines, calculate the intersection of the normal vector equation of the license plate to the road surface and the parking space direction equation, and determine whether the intersection solution is within the value range, so as to determine the relative position of the vehicle and the parking space.

[0007] Furthermore, in the step 1), the image reading and converting the RGB image into the HSV space specifically includes: collecting the image in the parking area through a camera as the input of the detection, and after the image is loaded, the matrix data of the workspace is returned; and the obtained RGB image is converted into the HSV space through a series of mathematical calculations.

[0008] Furthermore, in the step 2), the ground and the parking space wireframe are usually identified according to the obtained HSV component map, the approximate position of the ground and the parking space wireframe are determined, and finally they are replicated into the blank image.

[0009] Furthermore, in step 3), the parking space line extraction uses a "filter" with a certain range size to filter noise points, and the filter is traversed through all pixels of the image to finally obtain the pixel coordinates of the noise points. At the position of the obtained noise point coordinates, all pixels are assigned a value of 0. For the problem of wireframe holes or missing wireframes, morphological reconstruction technology is used to supplement them. Finally, the approximate part of the parking space line is reproduced.

[0010] Furthermore, in the step 3), obtaining the parking space line contour includes: calling the Canny edge detection algorithm to accurately detect the image edge through Gaussian blur denoising, calculating gradient amplitude and direction, non-maximum suppression, dual threshold detection and hysteresis threshold.

[0011] Furthermore, the step 4) includes intercepting the ground contour and extracting the ground contour line.

[0012] Furthermore, in step 4), the specific steps of calculating the close parts of the ground contour and the parking space contour, extracting the pixel point coordinates, and obtaining the parking space contour adjusted by the ground contour are: comparing the coordinates of the ground contour and the coordinates of the parking space contour in pairs, setting a threshold, traversing all the coordinate points of the two, and if the pixel coordinate area of ​​the two exceeds the threshold, it is determined to be not close, otherwise it is determined to be close, and the parking space line frame contour coordinates close to the ground contour coordinates are screened out to obtain the adjusted parking space line contour. After screening, the coordinates of the actual parking space contour will be mostly retained, and some larger noise blocks that were not removed when removing noise points before can be removed in this step.

[0013] Furthermore, the step 5) specifically includes: searching for the license plate position in the image through a "license plate search algorithm", taking a screenshot of its image and pixel coordinates, and marking it in the parking space line contour image of step 4) according to its pixel coordinates.

[0014] Furthermore, the step 6) specifically includes: performing cluster analysis on the contour line to obtain the category of each part of the contour line. Specifically, it includes: clustering the coordinates of the parking space wireframe using a Gaussian mixture clustering algorithm (GMM algorithm) to distinguish different sides of the parking space wireframe.

[0015] Furthermore, the step 6) also includes: for the edges of different parking space frames after clustering, solving their slope and intercept coefficient according to linear programming, representing the direction of the wireframe by a straight line, and taking the average direction of the wireframe as the parking space direction to obtain the parking space direction equation; the wireframe size of the parking space contains value range information; making a normal vector equation of the license plate to the road surface through the license plate direction to the wireframe; solving the equation group consisting of the normal vector equation of the license plate to the road surface and the parking space direction equation; if the equation group has an intersection solution and the intersection solution is within the value range, the vehicle stays in the parking space, otherwise it is outside the parking space, thereby determining the relative position of the vehicle and the parking space.

[0016] Based on the image shooting detection method, the present invention uses the intersection of the normal vector equation of the license plate to the road surface and the parking space direction equation to determine the relative position of the vehicle and the parking space, which effectively reduces the influence of adverse factors such as different vehicle models, lighting conditions, and obstructions, and effectively reduces the computational complexity of determining the parking position analysis of roadside vehicles, and greatly reduces the demand for computing resources. The present invention focuses on the recognition of the relative relationship between parking spaces and vehicles, and is used to determine whether the vehicle stays in the parking space under various abnormal circumstances, thereby providing services for vehicle parking billing, illegal parking, and abnormal state processing.

[0017] In a second aspect, the present invention further provides a system for identifying irregular road parking, which determines a parking space and a relative parking position based on the aforementioned irregular road parking identification method.

[0018] In a third aspect, the present invention further provides a computer device, comprising a processor and a memory, wherein the processor is connected to the memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory so that the computer device performs the steps of implementing the above method.

[0019] In a fourth aspect, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the steps of the above method when executed by a processor.

[0020] Beneficial Effects By implementing the above-mentioned road irregular parking recognition system provided by the present invention, the following technical effects are achieved: (1) Based on the image capture detection method, the method provided by this technical solution uses the intersection of the normal vector equation of the license plate to the road surface and the parking space direction equation to determine the relative position of the vehicle and the parking space, effectively reducing the influence of adverse factors such as different vehicle models, lighting conditions, and obstructions, effectively reducing the computational complexity of roadside vehicle parking position analysis, and greatly reducing the degree of computing resource requirements; (2) The method provided by this technical solution can realize efficient parking space detection and retrieval, effectively realizing the maximization of parking space resource utilization; (3) The method provided by the present technical solution does not require the installation of additional equipment for parking spaces where roadside parking vehicles are located, thereby reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to make the above-mentioned image recognition algorithm and computer vision algorithm of the present invention more obvious and easy to understand, the drawings required for use in the specific implementation of the present invention are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without paying creative labor.

[0022] Figure 1 Represents the workspace variable for loading images; Figure 2 Represents a graphics window showing a loaded image; Figure 3 Indicates conversion from RGB to HSV; Figure 4 Indicates the approximate identification of the parking space wireframe; Figure 5Indicates noise removal; Figure 6 It means morphological reconstruction and superposition to remove noise; Figure 7 Represents the outline of the parking space wireframe extracted by the Canny algorithm; Figure 8 It indicates the interception of the ground; Fig. 9 Represents the extraction of ground contour lines; Fig.10 The coordinates of the parking space outline frame close to the ground contour line; Fig.11 Represents the contour line of the parking space wireframe coordinates after filtering with the ground contour line coordinates; Fig.12 Indicates a screenshot of a license plate; Fig.13 Indicates the relative position of the license plate in the parking space wireframe outline; Fig.14 Cluster analysis of parking space wireframes; Fig.15 Linear program representing the parking space wireframe and license plate location. DETAILED DESCRIPTION

[0023] The present invention will be further described below in conjunction with the accompanying drawings.

[0024] The present invention discloses a method for identifying irregular parking on a road, comprising the steps of collecting an image in a parking area through a camera, reading the image and converting the RGB image into an HSV space, identifying the current status of the ground and the parking space according to the obtained HSV component map, determining the approximate location of the ground and the current status of the parking space, and finally replicating them into a blank image; for parking space line extraction, setting a "filter point" is used to remove noise points, and using morphological reconstruction to supplement the holes of the wireframe to obtain the parking space line contour line; calculating the close part of the ground and the parking space contour line, extracting the pixel point coordinates, and obtaining the parking space line contour line adjusted by the ground contour line; searching for the license plate position in the image through a "license plate search algorithm", and marking it in the parking space line contour line image; clustering analysis of the contour line, calculating the approximate directions of each part of the contour line, and planning the location of the parking space; and finally judging the relative position of the vehicle and the parking space by judging whether the equation group consisting of the normal vector ray equation and the parking space direction equation has a solution within a value range.

[0025] The method for identifying irregular parking on a road provided by the present invention specifically comprises the following steps: 1) Read the image and convert the RGB image to HSV space.

[0026] In some specific embodiments, the algorithm program developed based on Matlab software needs to be run based on the Matlab software platform, and can also be packaged as .exe software and run independently via the Matlab program, or can be loaded and run by other programming languages ​​such as Python through API interfaces. To load an image in the Matlab program, a specified image format is required, such as 'jpg', 'png', 'tif', etc. However, in general, Matlab can automatically determine the correct format based on the file extension. It is also possible to specify the frame to be read for an image containing multiple pages (such as a GIF or TIFF file with multiple frames).

[0027] The present invention uses a camera to collect images in the parking area as the input for detection. After the image is loaded, the matrix data of the workspace will be returned. For color images, an M×N×3 matrix is ​​returned, in which the third dimension is used to store the red, green and blue (RGB) color channels. Figure 1 As shown, Figure 1 Represents color image matrix data.

[0028] RGB color space: RGB stands for the three primary colors red, green and blue. RGB images use the color information of the three channels of red, green and blue to describe each pixel in the image. The value of each channel is usually represented by an integer from 0 to 255. By adjusting the values ​​of these three channels, various colors can be mixed. The RGB color space is a hardware-oriented color model that is widely used in computer graphics and image processing.

[0029] HSV color space: HSV is the initials of Hue, Saturation, and Value. The HSV model is described based on the intuitive characteristics of color, which is more in line with the visual perception of the human eye. Among them, hue represents the basic attributes of color, such as red, green, etc.; saturation represents the purity or vividness of the color; and value represents the brightness of the color. HSV color space is usually used for tasks such as image segmentation, color recognition, and color adjustment.

[0030] Compared with RGB images, HSV color space is more suitable for tasks such as image segmentation, color recognition and color adjustment. Therefore, the present invention needs to convert RGB images into HSV color space. RGB and HSV are two different color spaces, which describe colors in different ways. Converting RGB images to HSV space is the process of converting images from RGB color space to HSV color space.

[0031] Converting an RGB image to the HSV color space requires the following steps: Calculate the maximum, minimum, and difference values ​​in the three RGB channels; Determine the hue, saturation, and brightness in the HSV space based on the relationship between the difference and the maximum / minimum values; Normalize the calculated HSV values ​​to make them conform to the value range of the HSV color space. Through this conversion process, each pixel in the RGB image is mapped to the corresponding position in the HSV color space, thereby realizing the conversion of the RGB image to the HSV space. This conversion helps to extract color features in the image, which is convenient for subsequent image processing and analysis tasks. Figure 3 As shown, Figure 3 Represents the HSV component map, where Hue, Saturation, and Color Value represent hue, saturation, and color brightness, respectively.

[0032] 2) Process the HSV space and replicate the approximate position of the parking space line.

[0033] according to Figure 3 The HSV component graph in the figure determines the recognition of different colors of the ground and parking space status. It can be seen that in the recognition of the ground and parking space status, the H (Hue) dimension tends to be completely black for the ground, the S (Saturation) dimension is also completely black, and the V (Color Value) dimension has a clear difference between the ground and the parking space wireframe. In reality, the parking space wireframe must be on the road surface. The ground position is determined according to the H (Hue) and S (Saturation) dimensions, and the approximate position of the parking space status is filtered by the V (Color Value) dimension and reproduced in the blank image, such as Figure 4 shown.

[0034] 3) For parking space line extraction, remove noise, perform morphological reconstruction, and obtain the parking space line contour.

[0035] like Figure 4 As shown in the figure, after obtaining the binary image of the parking space wireframe, there are most of the noise in the background and road surface. Since in the previous step, we copied the recognized image to a completely white blank image, Figure 4 Most of the space in is completely white, that is, the value in the binarization is 0. The part with the parking space wireframe has a binarization value of 1, and the pixels are usually continuous. Compared with the noise, although the binary value of some noise points is also 1, the noise points are usually smaller, that is, the pixels are discontinuous.

[0036] First, remove the noise by filtering. Specifically, set a "filter" of a certain size to filter the noise. The specific method is: set a circular filter with a radius of 3-20 pixels (the parameter can be set). In the position covered by the circular filter, if more than 90% of the pixels are 1 (the parameter can be set), it is considered to be a parking space wireframe, and 0% of the pixels are 1, it is considered to be blank, otherwise it is considered to be noise, and the filter is traversed through the pixels of all images to finally obtain the pixel coordinates of the noise points. For the coordinates of the obtained noise point positions, all pixels are assigned 0. Figure 5 Represents an image with noise removed.

[0037] After filtering out the noise, there are still some problems of wireframe holes and missing parts. The present invention uses morphological reconstruction technology to supplement this problem. Morphological reconstruction is an image processing technology based on morphological processing. It involves two images and a structural element: one image is used as a marker and is the starting point of the transformation; the other image is used as a template to constrain the transformation process; and the structural element is used to define connectivity. The marker image (starting point) is repeatedly expanded until the outline of the marker image matches the template image. Morphological reconstruction includes two steps: seed point marking and expansion. The seed point refers to the position of the target we want to reconstruct in the image. These seed points can be selected manually or automatically obtained through an image segmentation algorithm. Morphological reconstruction uses the shape and structure information in the image to locate, segment and measure the target. In addition, the opening operation based on morphological reconstruction can accurately restore the shape of the object before corrosion and only retain the image of the target area. Figure 6 Represents the image after morphological reconstruction and superposition to remove noise.

[0038] at this time Figure 6 The wireframe graphics are relatively complete, and we can perform edge detection on the graphics and extract the contour wireframe. The present invention uses the classic Canny edge detection algorithm to perform edge detection on the graphics and extract the contour wireframe, which follows three main criteria: low error rate, accurate positioning and clear edge correspondence.

[0039] The specific steps of the Canny edge detection algorithm are: 1. Gaussian blur denoising. Edge detection algorithms are usually very sensitive to noise. Therefore, before edge detection, filtering and denoising are required. In the Canny algorithm, a Gaussian filter is used to smooth the image to reduce the noise in the image.

[0040] 2. Calculate the gradient magnitude and direction. The core of edge detection is to calculate the gradient magnitude and direction of each pixel in the image. The gradient direction indicates the direction in which the pixel gray value changes fastest, while the gradient magnitude indicates the degree of change in that direction. In the Canny algorithm, the Sobel operator and other methods can be used to calculate the gradient.

[0041] 3. Non-maximum suppression. After calculating the gradient magnitude and direction, the edges in the image may still be wide and need to be refined. Non-maximum suppression is a method that traverses each pixel in the gradient magnitude image. If the gradient magnitude of the pixel is not the largest in its gradient direction, it is set to zero. This retains the points with the largest local gradient, i.e., edge points.

[0042] 4. Double threshold detection. After non-maximum suppression, it is necessary to determine which edges are real edges and which are noise or weak edges. The Canny algorithm uses a double threshold method to achieve this. It sets two thresholds: a high threshold and a low threshold. If the gradient amplitude of a pixel is higher than the high threshold, it is marked as a strong edge pixel. If it is lower than the low threshold, it is suppressed. If it is between the two, it is marked as a weak edge pixel.

[0043] 5. Hysteresis threshold, for pixels marked as weak edges, if they are connected to strong edge pixels, they are retained as edges, otherwise they are suppressed, which ensures that the detected edges are continuous and removes isolated noise points.

[0044] The above steps of the Canny edge detection algorithm work together to ensure that the detected edges have low error rate, accurate positioning and clear response. Figure 7 As shown, the wireframe image is obtained after noise removal and morphological reconstruction.

[0045] 4) Calculate the approximate part of the ground contour and the parking space contour, extract the pixel point coordinates, and obtain the parking space contour adjusted by the ground contour.

[0046] like Figure 8 and Fig. 9 As shown in the figure, in order to solve the problem of the incompleteness of the parking space wireframe in the image due to years of wind and rain, it is necessary to extract the contour line of the ground part and determine the planning of the parking space wireframe by selecting the intersection of the ground and the parking space wireframe. Figure 7 The coordinates of the parking space wireframe contour extracted by the Canny algorithm and Fig. 9 The ground contour line coordinates extracted from the , set a suitable threshold, traverse all the coordinate points of the two, if the threshold is exceeded, it is judged as not close, otherwise it is judged as close. The selected parking space wireframe contour line coordinates close to the ground contour line coordinates are as follows Fig.10 After screening, the coordinates of the actual parking space contour lines will be mostly retained, and the larger noise blocks that were not removed when removing noise points are removed in this step, as shown in Fig.11 shown.

[0047] 5) Search the license plate position in the image and mark it in the parking space line contour image of step 4). Fig.12 As shown, the license plate position is searched in the image by the "license plate search algorithm". The license plate search algorithm in the prior art can be used in the present invention, and the image and pixel coordinates are captured, and the pixel coordinates are marked to Fig.12 The result is as follows Fig.13 shown.

[0048] 6) By calculating the intersection of the normal vector equation of the license plate to the road surface and the parking space direction equation, the relative position of the vehicle and the parking space can be determined. Fig.14 As shown, the present invention uses a Gaussian mixture clustering algorithm to perform cluster analysis on the contour lines to obtain and distinguish the categories of various parts of the contour lines. The red and blue categories in the figure are the two sides of the contour lines respectively. For the edges of the photo frames of different parking spaces after clustering, their slopes and intercept coefficients are solved according to linear programming, and the direction of their wireframes is represented by a straight line. For example, the direction of the blue category is represented by a black line, and the direction of the red wireframe is represented by a green line. The average direction of the black and green lines is the orientation of the parking space, represented by a thin red line. The green far point represents the position of the license plate, and the normal vector ray is drawn through the license plate to the plane defined by the black and green lines. The light blue straight line is the normal vector ray. The wireframe size of the parking space contains value range information. After the aforementioned steps, we have obtained the equation of the light blue normal vector ray and the equation of the thin red line, as shown Fig.15 As shown, at this time, solve the two straight line equations. If the system of equations has an intersection solution and the intersection solution is within the value range, the vehicle stays in the parking space, otherwise it is outside the parking space.

[0049] At this point, the algorithm solves the position of the vehicle relative to the parking space.

[0050] The method provided by the present invention is based on the image shooting detection method. After a series of steps, the normal vector equation of the license plate to the road surface and the parking space direction equation are obtained. The relative position of the vehicle and the parking space is determined by solving the intersection solution of the equation group composed of the two equations, which effectively reduces the influence of adverse factors such as different vehicle models, lighting conditions, and obstructions, and effectively reduces the computational complexity of determining the parking position analysis of roadside vehicles. The demand for computing resources is greatly reduced, and efficient parking space detection and retrieval can be achieved, effectively realizing the maximization of parking space resource utilization. In the solution process, there is no need to install additional equipment for the parking space where the roadside parking vehicle is located, thereby reducing costs.

[0051] In another embodiment, the present invention further provides a system for identifying irregular road parking, which determines a roadside parking space based on the aforementioned irregular road parking identification method.

[0052] In another embodiment, the present invention further provides a computer device, including a processor and a memory, wherein the processor is connected to the memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the computer device performs the steps of implementing the above method.

[0053] In another embodiment, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

Claims

1. A method for identifying irregular parking on a road, characterized in that: The following steps are involved: 1) Read the image and convert the RGB image to HSV space; 2) Process the HSV space and reproduce the approximate position of the parking space line; 3) For parking space line extraction, remove noise, perform morphological reconstruction, and obtain the parking space line contour; 4) Calculate the close parts of the ground contour line and the parking space line contour line, extract the pixel point coordinates, and obtain the parking space line contour line adjusted by the ground contour line; 5) Search the license plate position in the image and mark it in the parking space line contour image of step 4); 6) Perform cluster analysis on the contour lines, calculate the intersection of the normal vector equation of the license plate to the road surface and the parking space direction equation, and determine whether the intersection solution is within the value range, so as to determine the relative position of the vehicle and the parking space.

2. The method for identifying irregular parking on a road according to claim 1, characterized in that: In the step 1), the image reading and converting the RGB image into the HSV space specifically includes: collecting the image in the parking area through a camera as the input of the detection, and after the image is loaded, the matrix data of the workspace is returned; and the obtained RGB image is converted into the HSV space through a series of mathematical calculations.

3. The method for identifying irregular parking on a road according to claim 2, characterized in that: In the step 2), the ground and the parking space wireframe are usually identified according to the obtained HSV component map, the approximate position of the ground and the parking space wireframe are determined, and finally they are replicated into the blank image.

4. The method for identifying irregular parking on a road according to claim 3, characterized in that: In the step 3), obtaining the parking space line contour includes: calling the Canny edge detection algorithm to accurately detect the image edge through Gaussian blur denoising, calculating gradient amplitude and direction, non-maximum suppression, dual threshold detection and hysteresis threshold.

5. The method for identifying irregular parking on a road according to claim 4, characterized in that: The step 4) includes intercepting the ground contour and extracting the ground contour line.

6. The method for identifying irregular parking on a road according to claim 5, characterized in that: The step 5) specifically includes: searching for the license plate position in the image by means of a "license plate search algorithm", taking a screenshot of the image and pixel coordinates thereof, and marking the image of the parking space contour line of step 4) according to the pixel coordinates thereof.

7. The method for identifying irregular parking on a road according to claim 6, characterized in that: The step 6) specifically includes: performing cluster analysis on the contour line to obtain the category of each part of the contour line. Specifically includes: using a Gaussian mixture clustering algorithm to cluster the coordinates of the parking space wireframe to distinguish different sides of the parking space wireframe.

8. The road irregular parking recognition system is characterized by: The system determines the parking space and the relative parking position based on the method described in any one of claims 1 to 7.

9. A computer device, characterized in that: The computer device comprises a processor and a memory, wherein the processor is connected to the memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the computer device executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 7 is implemented.

Images