A License Plate Auxiliary Recognition Method and System for Intelligent Checkpoints

Abstract

The present invention relates to the field of traffic control technology, and particularly relates to a license plate auxiliary recognition method and system for an intelligent checkpoint. In the method: the character to be processed is matched with a preset character library to obtain a matching character; then, according to the shape characteristics of the matching character, the surrounding area is grown to obtain a complete target character. This license plate auxiliary recognition method can cope with situations such as occlusion and damage of license plate characters, and improve the accuracy of character recognition. Through accurate license plate auxiliary recognition results, the intelligent checkpoint can accurately identify illegal vehicles, automatically record and trigger corresponding traffic control measures, such as fines and points deductions; the intelligent checkpoint can, through accurate license plate auxiliary recognition results, analyze traffic data in real time and accurately, provide timely and accurate information support for traffic management departments, help traffic management departments make decisions quickly, optimize traffic flow, and improve road traffic efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic control, and particularly to a license plate auxiliary recognition method and system for an intelligent checkpoint. Background Art

[0002] In scenarios such as traffic control, urban security monitoring, parking lot management, as well as complex scenarios such as intelligent construction sites and intelligent gas stations, intelligent checkpoints can be set up at important traffic intersections or sections. An intelligent checkpoint is a set of intelligent devices for monitoring and managing traffic flow. By setting up an intelligent checkpoint, not only can the basic vehicle passing record function be achieved, but also it can be combined with license plate recognition technology to provide richer services and applications using the results of license plate recognition.

[0003] However, when an intelligent checkpoint is combined with license plate recognition technology, license plate recognition is affected by factors such as environmental conditions, license plate styles, and device hardware performance, reducing the accuracy of license plate recognition results. When an intelligent checkpoint uses the license plate recognition results with reduced accuracy for traffic control, it will wrongly report illegal vehicles or allow illegal vehicles to pass, resulting in an increase in traffic management errors; it will cause the traffic flow data counted by the intelligent checkpoint to be inaccurate, thereby affecting the reliability of traffic flow analysis and traffic planning and scheduling decisions; it will cause traffic management measures (such as license plate restrictions for travel, bans during special periods, etc.) to be not implemented effectively, unable to effectively regulate the traffic flow, and worsening the traffic congestion situation; for vehicles that need special attention (such as dangerous goods transport vehicles, stolen vehicles, etc.), if the intelligent checkpoint cannot correctly identify them, these high-risk vehicles may be missed, increasing the risk of public safety.

[0004] Therefore, how to improve the accuracy of traffic control by an intelligent checkpoint is an urgent problem to be solved currently. Summary of the Invention

[0005] In order to solve the technical problem of how to improve the accuracy of traffic control by an intelligent checkpoint, the purpose of the present invention is to provide a license plate auxiliary recognition method and system for an intelligent checkpoint, and the specific technical solutions adopted are as follows:

[0006] An embodiment of the present application provides a license plate auxiliary recognition method for an intelligent checkpoint, and the method includes:

[0007] Obtain a front image of a vehicle collected by an intelligent checkpoint;

[0008] Perform contour detection on the front image of the vehicle, and obtain a license plate image according to the result of the contour detection, where the license plate image includes a plurality of characters to be processed;

[0009] For any of the to-be-processed characters, match the to-be-processed character with a preset character library to obtain a matching character, and perform region growing on the periphery of the to-be-processed character according to the shape feature of the matching character to obtain a target character;

[0010] Verify the target character according to the matching character. When the result of the verification meets a preset verification condition, use the target character as the result of license plate auxiliary recognition, and perform traffic control according to the result of the license plate auxiliary recognition.

[0011] In some embodiments, after obtaining the license plate image, it includes:

[0012] When the license plate image is in color, perform grayscale processing on the license plate image to obtain a grayscale image corresponding to the license plate image;

[0013] Perform dilation processing on the grayscale image through a preset structural element to obtain a dilated grayscale image;

[0014] Perform erosion processing on the dilated grayscale image through the preset structural element to obtain an eroded grayscale image;

[0015] Perform threshold segmentation processing on the eroded grayscale image to obtain a binary license plate image, and the binary license plate image is used to match with the preset character library.

[0016] In some embodiments, the step of matching the to-be-processed character with a preset character library to obtain a matching character includes:

[0017] Determine the preset character library corresponding to the to-be-processed character according to the position of the to-be-processed character, and calculate the matching degree between the to-be-processed character and multiple reference characters in the preset character library;

[0018] Determine the matching character from the multiple reference characters according to the matching degree.

[0019] In some embodiments, the step of calculating the matching degree between the to-be-processed character and multiple reference characters in the preset character library includes:

[0020] Perform corner detection on the to-be-processed character to obtain a first corner set corresponding to the to-be-processed character;

[0021] Perform corner detection on each of the reference characters to obtain a second corner set corresponding to each of the reference characters;

[0022] Map the positions of the corners in the first corner set to the second corner set, and determine the corners around the mapped positions in the second corner set that satisfy a preset distance threshold;

[0023] Obtain the matching degree between the character to be processed and each reference character according to the number of corner points around the mapped position in the second set of corner points that satisfy a preset distance threshold.

[0024] In some embodiments, the region growing around the character to be processed according to the shape characteristics of the matching character includes:

[0025] When there are corner points around the first corner point in the first set of corner points that satisfy the preset distance threshold when its position is mapped to the second set of corner points, use the first corner point as a seed point;

[0026] Set the gray value difference threshold when filling the pixel points around the seed point, and set the region growing direction according to the stroke characteristics corresponding to the position of the seed point in the matching character;

[0027] For any one of the seed points, perform region growing on the periphery of the seed point according to the gray value difference threshold and the region growing direction.

[0028] In some embodiments, the performing region growing on the periphery of the seed point according to the gray value difference threshold and the region growing direction includes:

[0029] Determine a plurality of pixel points from the periphery of the seed point according to the gray value difference threshold to obtain the region to be filled;

[0030] Perform gray filling on the pixel points in the region to be filled, and connect the seed point with the pixel points after gray filling to obtain a growth vector;

[0031] Determine the growth possibility according to the included angle between the region growing direction and the growth vector;

[0032] Compare the growth possibility with a preset growth possibility threshold, and control the region growing process according to the comparison result.

[0033] In some embodiments, the controlling the region growing process according to the comparison result includes:

[0034] When the comparison result is that the growth possibility is greater than or equal to the preset growth possibility threshold, use the pixel points after gray filling as new seed points, and perform region growing on the periphery of the new seed points according to the gray value difference threshold;

[0035] When the comparison result is that the growth possibility is less than the preset growth possibility threshold, stop performing region growing on the periphery of the seed point.

[0036] In some embodiments, after performing region growing around the new seed point, the following steps are further included:

[0037] Based on the new character region obtained by performing region growing around the new seed point, establish multiple edge vectors for the edge pixel points of the new character region along the region growing direction, and calculate the angular change difference between every two adjacent edge vectors;

[0038] According to the angular change difference, obtain the likelihood degree of the character pixel points corresponding to the edge pixel points;

[0039] Along the region growing direction, determine the horizontal distance between the spaced edge pixel points;

[0040] Combine the likelihood degree of the character pixel points and the horizontal distance to control the process of region growing.

[0041] In some embodiments, the preset verification condition includes a preset threshold for the coincidence degree of character inflection points. When verifying the target character according to the matching character and the verification result meets the preset verification condition, taking the target character as the result of license plate auxiliary recognition includes:

[0042] Compare the inflection points of the matching character with the inflection points of the target character to obtain the coincidence degree of character inflection points;

[0043] When the coincidence degree of character inflection points is greater than or equal to the preset threshold for the coincidence degree of character inflection points, determine that the target character is the result of license plate auxiliary recognition.

[0044] The embodiments of the present application further provide a license plate auxiliary recognition system for an intelligent checkpoint, and the system includes:

[0045] An image acquisition module, configured to acquire a front image of a vehicle collected by the intelligent checkpoint;

[0046] A contour detection module, configured to perform contour detection on the front image of the vehicle, and obtain a license plate image according to the result of the contour detection, where the license plate image includes a plurality of characters to be processed;

[0047] A region growing module, configured to match any one of the characters to be processed with a preset character library to obtain a matching character, and perform region growing around the character to be processed according to the shape characteristics of the matching character to obtain a target character;

[0048] The character verification module is used to verify the target character according to the matching character, and when the verification result meets the preset verification condition, the target character is used as the result of the license plate auxiliary recognition, and traffic control is performed according to the result of the license plate auxiliary recognition.

[0049] The present invention has the following beneficial effects:

[0050] First, obtain the front image of the vehicle collected by the smart card port; then, perform contour detection on the front image of the vehicle, and obtain a license plate image based on the result of the contour detection, wherein the license plate image includes a plurality of characters to be processed; then, for any of the characters to be processed, match the characters to be processed with a preset character library to obtain a matching character, and perform region growth around the characters to be processed based on the shape characteristics of the matching characters to obtain a target character; finally, verify the target character based on the matching characters, and when the result of the verification meets the preset verification conditions, use the target character as the result of license plate auxiliary recognition, and perform traffic control based on the result of the license plate auxiliary recognition. In the present application, the characters to be processed are matched with a preset character library to obtain matching characters; then, based on the shape characteristics of the matching characters, the surrounding area is grown to obtain a complete target character. The license plate auxiliary recognition method can cope with situations such as occlusion and contamination of license plate characters, and improve the accuracy of character recognition. Through accurate license plate auxiliary recognition results, the smart toll gate can accurately record and trigger corresponding traffic control measures such as fines and point deductions for illegal vehicles; the smart toll gate can accurately analyze traffic data in real time through accurate license plate auxiliary recognition results, and provide timely and accurate information support for traffic management departments, which helps traffic management departments make decisions quickly, optimize traffic flow, and improve road traffic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 A schematic diagram of an implementation environment of a smart card slot license plate auxiliary recognition method provided by an embodiment of the present invention;

[0053] Figure 2 A flowchart of a method for assisting license plate recognition of a smart card slot provided by an embodiment of the present invention;

[0054] Figure 3Schematic diagram of a binary license plate image provided by an embodiment of the present invention;

[0055] Figure 4 Schematic diagram of the structure of a license plate auxiliary recognition system for an intelligent checkpoint provided by an embodiment of the present invention. Detailed implementation manners

[0056] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features and effects of a license plate auxiliary recognition method and system for an intelligent checkpoint according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0057] It should be noted that the terms "first", "second", etc. in the specification of this application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this invention belongs.

[0059] The following specifically describes the specific solutions of a license plate auxiliary recognition method and system for an intelligent checkpoint provided by the present invention with reference to the accompanying drawings.

[0060] Please refer to Figure 1 , Figure 1 Schematic diagram of the implementation environment of a license plate auxiliary recognition method for an intelligent checkpoint provided by an embodiment of the present invention. As Figure 1As shown in the figure, the implementation environment includes a license plate auxiliary recognition terminal 101 and an image acquisition terminal 102. The license plate auxiliary recognition terminal 101 can be a terminal device configured with a license plate auxiliary recognition system, including but not limited to a laptop computer, a tablet computer, a personal digital assistant, a PAD (tablet computer), a desktop computer, etc. with local computing capabilities; the license plate auxiliary recognition system can be implemented in the form of a target client, and the target client can be a video client, an instant messaging client, a browser client, etc. that support license plate auxiliary recognition in intelligent checkpoints; the license plate auxiliary recognition terminal 101 can communicate with the image acquisition terminal 102 through a network, which can include but not limited to: a wired network and a wireless network. Among them, the wired network includes: a local area network, a metropolitan area network, and a wide area network, and the wireless network includes: Bluetooth, WIFI (Wireless Fidelity, a technology that allows electronic devices to connect to a wireless local area network), and other networks that implement wireless communication. The above license plate auxiliary recognition terminal 101 can include but not limited to a human-computer interaction screen, a processor, and a memory. The above human-computer interaction screen can be used to display the front image of the vehicle and the result of license plate auxiliary recognition. The above processor can be used to respond to human-computer interaction operations, execute corresponding operations, or generate corresponding instructions.

[0061] As an alternative, the image acquisition terminal 102 is set at the intelligent checkpoint. When a vehicle passes through the intelligent checkpoint, the front image of the vehicle can be acquired through the image acquisition terminal 102. Among them, the image acquisition terminal 102 can include, for example: an infrared camera, which is suitable for night or low-light environments and can provide clear images at night or under low-light conditions to enhance the all-weather working ability of the system; an ordinary camera, which is the most basic image acquisition device, suitable for most conventional scenarios, captures light through an optical lens, converts the light into an electrical signal, and then converts it into a digital image by an image sensor; a panoramic camera, which is used to provide a wide range of views and is suitable for scenarios that need to cover a wide area.

[0062] As an alternative, the above license plate auxiliary recognition terminal 101 can also be a server, which can be a single server, a server cluster composed of multiple servers, or a cloud server. The above is only an example, and no limitation is made in this embodiment.

[0063] As an alternative, the following steps of the license plate auxiliary recognition method for the intelligent checkpoint can be executed on the license plate auxiliary recognition terminal 101:

[0064] Obtain the front image of the vehicle collected at the intelligent checkpoint;

[0065] Performing contour detection on the front image of the vehicle, and obtaining a license plate image according to the result of the contour detection, wherein the license plate image includes a plurality of characters to be processed;

[0066] For any of the characters to be processed, the character to be processed is matched with a preset character library to obtain a matching character, and region growth is performed around the character to be processed according to shape features of the matching character to obtain a target character;

[0067] The target character is verified according to the matching character, and when the verification result meets the preset verification condition, the target character is used as the result of the license plate auxiliary recognition, and traffic control is performed according to the result of the license plate auxiliary recognition.

[0068] In the above method, the characters to be processed are matched with the preset character library to obtain matching characters; then, according to the shape characteristics of the matching characters, the surrounding area is grown to obtain the complete target characters. The license plate auxiliary recognition method can cope with the occlusion and contamination of license plate characters and improve the accuracy of character recognition. Through accurate license plate auxiliary recognition results, the smart card gate can accurately record and trigger corresponding traffic control measures such as fines and deductions for illegal vehicles; the smart card gate can accurately analyze traffic data in real time through accurate license plate auxiliary recognition results, and provide timely and accurate information support for traffic management departments, which helps traffic management departments make decisions quickly, optimize traffic flow, and improve road traffic efficiency.

[0069] As an optional example, this embodiment does not limit the execution subject of the above-mentioned smart card port license plate auxiliary recognition method. The above-mentioned smart card port license plate auxiliary recognition method can be executed on the license plate auxiliary recognition terminal 101. For example, when the license plate auxiliary recognition terminal 101 is a desktop computer, some or all steps of the above-mentioned smart card port license plate auxiliary recognition method can be executed on the desktop computer.

[0070] The above section introduces the contents of an exemplary implementation environment for applying the technical solution of the present application. Next, we will continue to introduce the license plate auxiliary recognition method of the smart checkpoint of the present application.

[0071] In order to solve the problem of how to improve the accuracy of traffic control by smart checkpoints in the prior art, the embodiments of the present application respectively propose a smart checkpoint license plate auxiliary recognition method and a smart checkpoint license plate auxiliary recognition system. These embodiments will be described in detail below.

[0072] See also Figure 2 , Figure 2 A flowchart of a smart card slot auxiliary license plate recognition method provided by an embodiment of the present invention is provided. The method can be applied to Figure 1The implementation environment shown. It should be understood that this method can also be applied to other exemplary implementation environments and be specifically executed by devices in other implementation environments. This embodiment does not limit the implementation environment applicable to this method.

[0073] As Figure 2 shown, in an exemplary embodiment, the license plate auxiliary recognition method for an intelligent checkpoint at least includes steps S210 to S240, which are introduced in detail as follows:

[0074] In step S210, obtain the frontal image of the vehicle collected by the intelligent checkpoint.

[0075] Among them, the frontal image of the vehicle refers to an image containing the front part of the vehicle captured by an image acquisition terminal (such as a camera), and this image can show the front structure of the vehicle. The front structure of the vehicle includes: the vehicle head, the front bumper, the headlights, etc.

[0076] Exemplarily, an intelligent checkpoint is set up on the main road of a city to monitor and manage traffic flow. The intelligent checkpoint is equipped with a high-definition camera installed on the columns on both sides of the road. The height of the high-definition camera is about 5 meters, and the shooting direction is perpendicular to the vehicle driving direction. When a vehicle enters the shooting range of the high-definition camera, the high-definition camera automatically starts the shooting function. The high-definition camera captures the frontal image of the vehicle, and the license plate, the vehicle head, the headlights, etc. are clearly shown in the frontal image of the vehicle.

[0077] In step S220, perform contour detection on the frontal image of the vehicle, and obtain a license plate image according to the result of the contour detection. The license plate image includes a plurality of characters to be processed.

[0078] Among them, contour detection is an image processing technology used to detect the boundaries or edges of objects in an image. When performing contour detection, the contour line of the object can be extracted by analyzing the pixel values in the image. Commonly used contour detection methods include Canny edge detection, Sobel operator, Laplacian operator, etc. Through contour detection, the license plate can be separated from the complex background of the frontal image of the vehicle.

[0079] Among them, the license plate image refers to a sub-image containing the license plate part extracted from the frontal image of the vehicle. The license plate image should contain the complete license plate.

[0080] Among them, the characters to be processed refer to the individual characters in the license plate image that need to be recognized. Each character is usually an independent area with a specific shape and size, and each character can be separated from the license plate image through character segmentation and recognition.

[0081] Exemplarily, after obtaining a front image of a vehicle, the front image of the vehicle is grayscaled to reduce the interference of color information; then, the Canny edge detection algorithm is used to perform contour detection on the image, extract the edge information in the image, and obtain multiple contours based on the edge information. Contours that may contain license plates are screened out from the multiple contours. Generally, a license plate contour has a certain aspect ratio and shape characteristics; then, according to the screened contours, the license plate area is extracted from the original image to obtain a license plate image; finally, the license plate image is further processed. For example, the characters in the license plate image are segmented by the horizontal projection method to obtain multiple characters to be processed.

[0082] In this embodiment, through contour detection, the license plate area can be accurately extracted, background interference can be reduced, and the accuracy of character recognition can be improved; contour detection can effectively filter out the edge information of non-license plate areas, reduce the false detection rate, and avoid misidentifying other objects as license plates; contour detection has good robustness to noise and slight deformation in the image and can adapt to different shooting conditions and environments.

[0083] In step S230, for any one of the characters to be processed, the character to be processed is matched with a preset character library to obtain a matching character, and region growing is performed around the character to be processed according to the shape characteristics of the matching character to obtain a target character.

[0084] Among them, the character library is a set containing various reference characters. The reference characters usually include letters, numbers, special symbols, etc. The preset character library is compared and matched with the character to be processed to determine the specific content of the character to be processed. The character library can be an image library or a feature vector library, depending on the matching algorithm used.

[0085] Among them, when matching the character to be processed with the preset character library, a preset matching algorithm can be used to compare the extracted character to be processed with the reference characters in the character library to find the most similar reference character as the matching character. The preset matching algorithms can include: template matching, feature point matching, convolutional neural network, etc.

[0086] Among them, the shape characteristics of the matching character refer to the geometric shape, stroke direction, stroke thickness, etc. of the matching character. Region growing is performed on the pixel points with similar gray values around the character to be processed using the shape characteristics of the matching character as the growth direction, and a judgment criterion for stopping region growing can also be obtained according to the shape characteristics of the matching character.

[0087] Among them, region growing is an image segmentation technique that gradually expands starting from seed points and merges pixels with similar attributes into the same region. When performing region growing, according to the set similarity criteria (such as color, texture, shape, etc.), starting from the seed points, the surrounding pixels can be gradually added to the target region. Through region growing, the boundaries of the characters to be processed can be repaired and extended to make them more complete and accurate.

[0088] Exemplarily, the character "B" to be processed is segmented from the license plate image; using the template matching algorithm, the character "B" to be processed is compared with the reference characters in the character library, the similarity between the character to be processed and each reference character is calculated, and the reference character with the highest similarity is selected as the matching character; the shape features of the matching character are extracted, such as stroke direction, stroke thickness, etc.; according to the shape features of the matching character, the similarity criteria (such as color, texture, shape, etc.) are set, and starting from the seed points of the character to be processed, the surrounding pixels with similar attributes are gradually added to the target character region; continue to expand until the target character region no longer changes or reaches a predetermined stop condition, and finally a more complete and accurate target character "B" is obtained.

[0089] In this embodiment, through character library matching, the specific content of the character to be processed can be accurately identified, reducing the possibility of misidentification; the region growing technique can repair the defects in the character image, such as broken strokes, noise points, etc., making the character more complete and clear; the shape features of the matching character provide reliable guidance for region growing, enhancing the adaptability of the system to different shooting conditions and image qualities.

[0090] In step S240, according to the matching character, the target character is verified. When the result of the verification meets the preset verification conditions, the target character is used as the result of license plate auxiliary recognition, and traffic control is performed according to the result of the license plate auxiliary recognition.

[0091] Among them, based on the matching character, further verification of the target character is performed to ensure its correctness and reliability. The verification methods can be, for example: checking whether the shape of the target character is consistent with the matching character, including edge direction, curvature, connectivity, etc.; checking whether the size of the target character is within a reasonable range to avoid characters that are too small or too large; checking the texture features of the target character, such as color, brightness, etc., to ensure consistency with the matching character; checking whether the position and arrangement of the target character in the entire license plate are reasonable to avoid isolated or misaligned characters.

[0092] Exemplarily, a certain intelligent checkpoint is located at an important intersection on the urban arterial road, responsible for monitoring traffic flow and identifying illegal vehicles. The intelligent checkpoint conducts the following traffic control based on the results of license plate assisted recognition: If the recognized license plate number matches the records in the illegal vehicle database, the system immediately issues an alarm; at the same time, the intelligent checkpoint can link with nearby traffic lights, temporarily change the signal timing, and guide the illegal vehicle into the checkpoint; notify the nearby traffic police, informing the specific location and detailed information of the illegal vehicle for timely interception and handling.

[0093] Exemplarily, a certain city has implemented a policy of restricting the use of vehicles based on the last digit of the license plate, prohibiting vehicles with certain last digits from entering the urban area during specific time periods. The intelligent checkpoint conducts the following traffic control based on the results of license plate assisted recognition: If the recognized license plate number falls within the restricted range, the system issues a warning and records the violation; the intelligent checkpoint can link with nearby electronic displays to show restricted use reminder information to remind drivers to comply with the regulations; when necessary, the system can control nearby traffic lights to restrict the passage of illegal vehicles and guide them out of the restricted area.

[0094] Exemplarily, a certain intelligent checkpoint is located near a hospital and needs to ensure that special vehicles such as ambulances can pass quickly. The intelligent checkpoint conducts the following traffic control based on the results of license plate assisted recognition: If the recognized license plate number belongs to a special vehicle, the system immediately initiates the priority passage procedure; control nearby traffic lights to provide a green light for the special vehicle to ensure its quick passage; when necessary, the system can coordinate the surrounding traffic flow to create road space for the special vehicle.

[0095] Exemplarily, a certain intelligent checkpoint is located at the entrance of a large shopping mall, responsible for managing the vehicles entering the parking lot. The intelligent checkpoint conducts the following traffic control based on the results of license plate assisted recognition: If the recognized license plate number belongs to a registered user, the system automatically opens the barrier at the entrance of the parking lot to allow the vehicle to enter; for non-registered users, the system can prompt the driver to make temporary parking payments or register as a member; record the vehicle entry time to provide a basis for subsequent fee calculation and management.

[0096] It can be seen from the above steps S210 to S240 that in the scheme proposed in this embodiment, the characters to be processed are matched with the preset character library to obtain matching characters; then, according to the shape characteristics of the matching characters, the surrounding area is grown to obtain the complete target characters. The license plate auxiliary recognition method can cope with the occlusion and contamination of license plate characters and improve the accuracy of character recognition. Through accurate license plate auxiliary recognition results, the smart card gate can accurately record and trigger corresponding traffic control measures such as fines and deductions for illegal vehicles; the smart card gate can accurately analyze traffic data in real time through accurate license plate auxiliary recognition results, provide timely and accurate information support for traffic management departments, and help traffic management departments make decisions quickly, optimize traffic flow, and improve road traffic efficiency.

[0097] In one embodiment of the present application, after obtaining the license plate image, the following steps are included:

[0098] When the license plate image is in color, grayscale processing is performed on the license plate image to obtain a grayscale image corresponding to the license plate image;

[0099] Performing dilation processing on the grayscale image by using a preset structural element to obtain a grayscale image after dilation processing;

[0100] Performing an erosion process on the grayscale image after the dilation process by using the preset structural element to obtain an eroded grayscale image;

[0101] The grayscale image after the corrosion process is subjected to threshold segmentation process to obtain a binary license plate image, and the binary license plate image is used for matching with the preset character library.

[0102] Among them, grayscale processing of the license plate image can convert the color image into a grayscale image, that is, each pixel only retains one intensity value (grayscale value). Commonly used methods include weighted average method, maximum value method, etc., which can reduce the interference of color information, simplify the image processing process, and improve the efficiency and accuracy of subsequent processing.

[0103] Among them, the structural element refers to a small image matrix used for image morphological processing, usually a 3x3 or 5x5 matrix. The structural element can be used for dilation and erosion processing to control the effect and range of the processing. Common structural elements include circle, square, cross, etc.

[0104] Among them, dilation processing refers to a morphological operation that expands the white regions in an image by convolving a structuring element with each pixel in the image. For each pixel, if any pixel in the structuring element intersects with the corresponding pixel in the image and is white, then that pixel becomes white. Through dilation processing, small holes in the image can be filled, adjacent white regions can be connected, and the connectivity of the target region can be enhanced.

[0105] Among them, erosion processing is also a morphological operation that shrinks the white regions in an image by convolving a structuring element with each pixel in the image. For each pixel, if all pixels in the structuring element intersect with the corresponding pixel in the image and are white, then that pixel remains white; otherwise, it becomes black. Through erosion processing, small noise points in the image can be removed, the boundaries of the target region can be refined, and the compactness of the target region can be enhanced.

[0106] Among them, threshold segmentation is an image processing technique that divides the pixels in an image into two parts, foreground (white) and background (black), by setting a threshold. Common threshold segmentation methods include global thresholding, local thresholding, etc. Through threshold segmentation, a grayscale image can be converted into a binary image, separating the target region (license plate characters) from the background region (non-character part), which is convenient for subsequent character recognition and processing.

[0107] Exemplarily, refer to Figure 3 , Figure 3 which is a schematic diagram of a binary license plate image provided by an embodiment of the present invention. As can be seen from Figure 3 , in the binary license plate image, the white part is the target region (i.e., the license plate characters), and the black part is the background region (i.e., the non-character part).

[0108] In this embodiment, through grayscale processing, dilation processing, erosion processing, and threshold segmentation processing, noise can be effectively removed, holes can be filled, boundaries can be refined, making the license plate characters clearer and more regular, and improving the accuracy of character recognition.

[0109] In an embodiment of the present application, the step of matching the character to be processed with a preset character library to obtain a matching character includes:

[0110] Determining the preset character library corresponding to the character to be processed according to the position of the character to be processed, and calculating the matching degree between the character to be processed and multiple reference characters in the preset character library;

[0111] Determining the matching character from the multiple reference characters according to the matching degree.

[0112] Among them, according to the position of the character to be processed in the license plate image, a suitable preset character library can be selected for matching. Characters in different positions may have different fonts, sizes, and styles. By preparing multiple character libraries in advance, with each character library corresponding to a specific position, the matching accuracy can be improved, and misrecognition caused by unmatched character libraries can be reduced.

[0113] Among them, when calculating the matching degree between the character to be processed and multiple reference characters in the preset character library, the similarity between the character to be processed and each reference character in the character library can be calculated through a preset matching algorithm. Commonly used matching algorithms include template matching, feature point matching, convolutional neural networks, etc. Through the matching degree, it can be determined which reference character in the character library is most similar to the character to be processed, so as to find the best matching character.

[0114] Among them, when determining the matching character from multiple reference characters according to the matching degree, the reference character with the highest calculated matching degree can be selected as the final matching character according to the calculated matching degree. Usually, the reference character with the highest matching degree is selected, or a threshold is set, and only the reference characters with a matching degree exceeding the threshold are considered valid matching characters.

[0115] Exemplarily, there are fixed regulations for the gaps between characters in a license plate, and the characters in the license plate are horizontally arranged. Based on the binary license plate image, the regions of the characters in the license plate are divided; for the divided characters, the character library is matched in combination with their positions to determine the matching characters. For the characters on the license plate number, due to features such as stroke order, different characters have different pixel point trends in different regions, which can be reflected by the position and number of corner points. At the same time, the distribution characteristics of the corner points can also be used as a criterion for judging whether two characters are the same character.

[0116] In this embodiment, selecting a suitable character library according to the character position can reduce misrecognition caused by unmatched character libraries and improve the accuracy of character recognition; calculating the matching degree and setting a threshold can effectively filter out reference characters with low similarity and reduce the possibility of misrecognition.

[0117] In an embodiment of the present application, calculating the matching degree between the character to be processed and multiple reference characters in the preset character library includes:

[0118] Performing corner point detection on the character to be processed to obtain a first set of corner points corresponding to the character to be processed;

[0119] Performing corner point detection on each of the reference characters to obtain a second set of corner points corresponding to each of the reference characters;

[0120] Map the positions of the corner points in the first set of corner points to the second set of corner points, and determine the corner points in the second set of corner points that satisfy a preset distance threshold around the mapped positions;

[0121] Obtain the matching degree between the character to be processed and each of the reference characters according to the number of corner points in the second set of corner points that satisfy the preset distance threshold around the mapped positions.

[0122] Among them, corner point detection is an image processing technology used to detect corner points or feature points in an image. Corner points are usually points with significant gray level changes in the image, which can reflect the local features of the image. By detecting corner points, the key feature points of the character to be processed and the reference characters are extracted, providing a basis for subsequent matching.

[0123] Among them, mapping the corner point positions of the character to be processed to the corner point positions of the reference characters can establish the corresponding relationship between the two. Geometric transformation methods such as affine transformation or perspective transformation can be used to map the corner point positions of the character to be processed to the corner point positions of the reference characters. Through position mapping, the feature points of the character to be processed can be aligned with the feature points of the reference characters, facilitating subsequent matching and comparison.

[0124] Among them, in the set of corner points of the reference character, find the corner points around the mapped position of the character to be processed, and the distance between these corner points and the mapped position is less than the preset distance threshold. Specifically, when implementing, distance metrics such as Euclidean distance or Manhattan distance can be calculated to screen out the corner points that meet the conditions. By screening the corner points that meet the distance threshold, the similarity degree between the character to be processed and the reference character can be determined.

[0125] Among them, when calculating the matching degree between the character to be processed and the reference character according to the number of corner points that meet the preset distance threshold, the number of corner points that meet the conditions can be used as a measure of the matching degree. The more the number, the higher the matching degree. By calculating the matching degree, the similarity between the character to be processed and the reference character can be determined, and the most similar reference character can be selected as the matching result.

[0126] Exemplarily, the license plate is divided into multiple character regions. There are differences in the character types and directions represented by different character regions, so the contents of the characters in the character libraries of different character regions will also be different. For example: In a common license plate, the first character from left to right is composed of Chinese characters, which are often used to represent the province; the second character is a capital English letter used to represent the prefecture-level city. To achieve character matching, first, the license plate region in the binary license plate image needs to be divided, and character library matching is performed according to the divided regions. They are evenly divided into several regions according to the number of characters by the fixed width and spacing of each character in the license plate. The reference characters in the character library corresponding to the characters at fixed positions on the license plate all have their fixed shapes, so the number and distribution positions of the corner points of these reference characters have their own characteristics. The number and positions of the corner points detected from a to-be-processed character in the binary license plate image are used to match the reference characters in the preset character library corresponding to this to-be-processed character.

[0127] Exemplarily, select the leftmost first to-be-processed character among the multiple to-be-processed characters segmented. Use the corner detection algorithm to detect the corner points of the to-be-processed character, and obtain the first corner point set A. Perform corner detection on all reference characters once to obtain multiple second corner point sets B. Match the first corner point set A with the multiple second corner point sets B. This matching has the characteristic of one-to-many. Then, arbitrarily select a set from the multiple second corner point sets B for matching, that is, perform matching operations on the corner points in the two sets. To determine which of the multiple second corner point sets B has the highest matching degree with the first corner point set A, that is, based on the corner points and their positions in the first corner point set A, respectively correspond to the corner points in each second corner point set B, obtain the matching degree corresponding to each second corner point set B, and select the character represented by the second corner point set B with a high matching degree as the matching character. The Euclidean distance can be used to find the corner point with the highest matching degree at this position in the second corner point set B based on the corner points in the first corner point set A. If the distance between a corner point in the first corner point set A and the corner point at this position in the second corner point set B is less than the preset distance threshold, it is considered that these two corner points in the first corner point set A and the second corner point set B are successfully matched. If there are multiple intersection points in the second corner point set B that all meet the preset distance threshold, then select the corner point with the smallest distance for matching. Perform character library matching on the first corner point set A. Here, one first corner point set A corresponding to one second corner point set B will obtain a matching degree, and there are multiple second corner point sets B in the character library, so there will be multiple matching degrees. Select the character corresponding to the maximum value among these matching degrees as the matching character.

[0128] In this embodiment, through corner detection and position mapping, key feature points of the character to be processed and the reference character can be extracted and aligned, improving the accuracy of matching; by determining the number of corner points that meet the preset distance threshold, reference characters with relatively low similarity can be effectively filtered out, reducing the possibility of misrecognition.

[0129] In an embodiment of the present application, the growing the area around the character to be processed according to the shape feature of the matched character includes:

[0130] When the first corner point in the first corner point set is position-mapped to the second corner point set and there are corner points around it that meet the preset distance threshold, the first corner point is used as a seed point;

[0131] Set the gray value difference threshold when filling the pixel points around the seed point, and set the area growth direction according to the stroke feature corresponding to the position of the seed point in the matched character.

[0132] For any one of the seed points, perform area growth around the seed point according to the gray value difference threshold and the area growth direction.

[0133] Among them, the seed point refers to the starting point for area growth, usually a certain key point or feature point in the image. When the first corner point in the first corner point set is position-mapped to the second corner point set, if there are corner points around it that meet the preset distance threshold, then this first corner point is used as the seed point. Selecting a suitable seed point ensures that the area growth starts from the correct starting point, improving the accuracy and reliability of the growth.

[0134] Among them, the gray value difference threshold refers to the maximum allowable value of the gray value difference between the pixel points newly added to the area and the seed point during the area growth process. According to the actual requirements and image characteristics, a suitable gray value difference threshold is set, usually an empirical value or determined through experiments. By setting the gray value difference threshold, it is ensured that the pixel points newly added to the area have similar gray values to the seed point, avoiding mis-adding background or noise points to the area and improving the accuracy of area growth.

[0135] Among them, the area growth direction refers to the direction of preferential expansion during the area growth process, usually determined according to the stroke features of the character. According to the position of the seed point in the matched character, analyze the stroke features (such as the direction and thickness of the stroke) at this position, and set the corresponding area growth direction. By setting a reasonable area growth direction, it is ensured that the area growth proceeds along the stroke direction of the character, avoiding deviating from the boundary of the character and improving the accuracy and coherence of the area growth.

[0136] Exemplarily, for region growing around the character to be processed, the position of the seed point needs to be determined first. The determination of the seed point position during region growing also determines the region growing direction and will affect the growing quality. Observe Figure 3 the binary license plate image shown in Figure 3 and it can be found that there are defects in the character edges formed by processing the license plate characters, and there are discontinuous situations. Then, in this case, the seed point needs to be set on the region, and the region around the seed point is filled through region growing. For the selection of the seed point, randomly select one of the above-mentioned successfully matched corner points and use it as the seed point.

[0137] Exemplarily, the pixel point gray values in the character region of the binary license plate image are relatively large, and along the stroke direction of the character, the gray value sizes are relatively similar and have extensibility. The standard for region growing is the approximation standard of gray values. Set a gray value difference threshold of 20. Based on the determined seed point, search for pixel points with a gray value difference within 20 around the seed point and fill them. For these filled pixel points, further judge the gray value difference of the surrounding pixel points.

[0138] Exemplarily, during the process of growing the seed point, filling all the pixel points that meet the conditions in all directions around the seed point and continuously growing will include the pixel points in other snowflake regions with similar gray values that are not in the character region, which will cause errors in license plate character recognition. In response to this situation, based on the shape characteristics of the matching character, the growing direction of the position where the seed point is located is determined. In the character library, locate the position point d corresponding to the selected seed point in the binary license plate image, locate the stroke position of d in the character, and determine the corresponding growing direction according to the stroke characteristics.

[0139] In this embodiment, by selecting a suitable seed point and setting a reasonable gray value difference threshold, it can be ensured that the region growing starts from the correct starting point and proceeds along the stroke direction of the character, improving the accuracy of character recognition. The gray value difference threshold can effectively filter out the background and noise points, avoid mis-adding irrelevant pixel points to the target region, and reduce the possibility of mis-recognition; the setting of the region growing direction is based on the stroke characteristics of the character, enhancing the adaptability to different character shapes and styles and improving the robustness of the system.

[0140] In an embodiment of the present application, the region growing around the seed point according to the gray value difference threshold and the region growing direction includes:

[0141] Determine a plurality of pixel points from around the seed point according to the gray value difference threshold to obtain a region to be filled;

[0142] Gray-fill the pixel points in the area to be filled, and connect the seed point to the pixel points after gray filling to obtain a growth vector;

[0143] Determine the growth possibility according to the angle between the region growth direction and the growth vector;

[0144] Compare the growth possibility with a preset growth possibility threshold, and control the region growth process according to the comparison result.

[0145] Among them, multiple pixel points are selected from the neighborhood around the seed point to form an area to be filled. Usually, 8 adjacent pixel points (4 direct adjacencies and 4 diagonal adjacencies) around the seed point are selected to determine the pixel points that may be filled, preparing for subsequent gray filling and region growth.

[0146] Among them, when gray-filling the pixel points in the area to be filled, set the gray value of the pixel points in the area to be filled to be the same as the gray value of the seed point. Traverse each pixel point in the area to be filled and set its gray value to the gray value of the seed point to ensure that the pixel points in the area to be filled have a similar gray value to the seed point, providing a consistent basis for region growth.

[0147] Among them, when connecting the seed point to the pixel points after gray filling to obtain a growth vector, draw a vector from the seed point to the pixel points after gray filling, which is called a growth vector, and calculate the vector between the seed point and each pixel point after gray filling. Through the growth vector, the direction and path of region growth can be determined.

[0148] Among them, when determining the growth possibility according to the angle between the region growth direction and the growth vector, the angle between the region growth direction and the growth vector can be calculated, and the growth possibility can be determined according to the size of the angle. Use the dot product or cross product of vectors to calculate the angle. The smaller the angle, the greater the growth possibility. Through the calculation of the angle, ensure that the region growth proceeds along the predetermined stroke direction and avoid deviating from the boundary of the character.

[0149] Exemplarily, based on the determined seed point, check the area to be filled where the difference in gray value between each pixel point in its 3×3 neighborhood and its gray value is not more than 20. Connect the pixel points in the area to be filled to the seed point and generate multiple vectors. Calculate the angles between these vectors and the growth direction Calculate the growth possibility of whether the pixel points around the seed point corresponding to the vector are suitable for continued growth through the angle. The representation of the growth possibility can be:

[0150]

[0151] Among them, f represents the growth possibility, The smaller it is, the greater the possibility that the pixel points around the seed point are suitable for continued growth. The larger f is, the greater the possibility that it is suitable for continued growth based on this pixel point.

[0152] In this embodiment, by determining the area to be filled, gray filling, generating a growth vector, and calculating the included angle, it can be ensured that the region growth proceeds along the predetermined stroke direction, improving the accuracy of character recognition; by setting a growth possibility threshold, background and noise points can be effectively filtered out, avoiding mis-adding irrelevant pixel points to the target area and reducing the possibility of mis-recognition; the setting of the region growth direction is based on the stroke characteristics of the character, enhancing the adaptability to different character shapes and styles and improving the robustness of the system.

[0153] In an embodiment of the present application, controlling the region growth process according to the comparison result includes:

[0154] When the comparison result is that the growth possibility is greater than or equal to the preset growth possibility threshold, the pixel point after gray filling is used as a new seed point, and region growth is performed around the new seed point according to the gray value difference threshold.

[0155] When the comparison result is that the growth possibility is less than the preset growth possibility threshold, stop performing region growth around the seed point.

[0156] Exemplarily, the preset growth possibility threshold is 0.85. If the growth possibility is greater than or equal to 0.85, the pixel point after gray filling is used as a new seed point, and region growth is performed around the new seed point according to the gray value difference threshold.

[0157] In an embodiment of the present application, after performing region growth around the new seed point, it further includes:

[0158] Based on the new character region obtained by performing region growth around the new seed point, establish multiple edge vectors for the edge pixel points of the new character region along the region growth direction, and calculate the angle change difference between every two adjacent edge vectors.

[0159] According to the angle change difference, obtain the possible degree of the character pixel point corresponding to the edge pixel point.

[0160] Along the region growth direction, determine the horizontal distance between the spaced edge pixel points.

[0161] Combining the possible degree of the character pixel point and the horizontal distance, control the process of the region growth.

[0162] Among them, the edge pixel points refer to the pixel points on the boundary of the character area, which usually have obvious gray-scale changes. Through the edge detection algorithm, the edge pixel points of the character area are extracted.

[0163] Among them, when establishing multiple edge vectors along the area growth direction, starting from the edge pixel points, multiple edge vectors are established along the area growth direction. For each edge pixel point, according to the area growth direction, a vector pointing to the inside of the character is drawn.

[0164] Among them, calculate the angle between every two adjacent edge vectors to obtain the angle change difference. Use the dot product or cross product of vectors to calculate the angle, and the angle change difference reflects the shape change between the edge pixel points. Through the angle change difference, evaluate the shape characteristics of the edge pixel points and determine the possibility of them being character pixel points.

[0165] Among them, in the area growth direction, calculate the horizontal distance between the spaced edge pixel points. Along the area growth direction, measure the horizontal distance between each pair of adjacent edge pixel points. Through the horizontal distance, evaluate the spatial relationship between the edge pixel points to ensure the coherence and rationality of area growth.

[0166] Among them, according to the angle change difference and horizontal distance of the edge pixel points, comprehensively evaluate the possibility of character pixel points and control the area growth process accordingly. Thresholds can be set. If the possibility of character pixel points and the horizontal distance meet the preset conditions, continue the area growth; otherwise, stop growing. Through comprehensive evaluation, ensure that the area growth proceeds along the stroke direction of the character, avoiding misgrowth and noise interference.

[0167] Exemplarily, the matching characters all have certain shape characteristics. The edge area of the matching characters is relatively smooth, and the strokes of the matching characters have a fixed width. However, the above steps only use the gray-scale value difference threshold and the growth direction as the growth conditions, which will have problems of incomplete growth or overgrowth. Therefore, during the growth process, for the new character area obtained by performing area growth around the new seed points, establish multiple edge vectors for the edge pixel points of the new character area along the area growth direction, and calculate the angle change difference between every two adjacent edge vectors. Average the multiple angle change differences to obtain the average value α of the angle change difference. Take this average value of the angle change difference as the judgment criterion, and judge the difference β between the angle between the vector generated between the edge pixel points of the filled area in the continuous growth process and the previous edge pixel point and the previous angle. The representation method of the possibility degree of character pixel points can be:

[0168]

[0169] Among them, j represents the possibility degree of character pixel points; ∣α―β∣ represents the difference between the included angle between two adjacent pixel point vectors at the edge of the latest growing region and the average value of the angle change difference during the growth process. The smaller this difference is, the more uniform the angle change of the edge pixel points of the new character region is.

[0170] Among them, the larger j is, the greater the possibility that the edge pixel point is a pixel point.

[0171] Exemplarily, the threshold of the possibility degree of character pixel points can be set to 0.85. If j is greater than or equal to 0.85, it is considered that the possibility that the edge pixel point is in the internal region of the character is greater, and at this moment, it is suitable to continue growing. If j is less than 0.85, then the growth stops at this moment.

[0172] Exemplarily, when using the smooth characteristics around the character as the condition for judging the stop of region growth, it includes regions with smooth edge contours but fat character shapes formed due to overgrowth. And the strokes of the characters on the license plate have a fixed width. Therefore, every k pixel points along the growth direction, two edge pixel points are located, perpendiculars are drawn from the two edge pixel points to the growth direction, and the horizontal distance D between the edge pixel points is obtained. The preset standard horizontal distance is L, and the difference m between the horizontal distance D between the edge pixel points and the standard horizontal distance is used to control the process of region growth. For example, the difference threshold is set to 0.1. During the growth process, if the m value is greater than or equal to 0.01, it is considered that the horizontal distance between the edge pixel points meets the reference character standard, and the growth can stop.

[0173] Among them, the growth satisfaction degree can be represented by the difference between the horizontal distance D between the edge pixel points and the standard horizontal distance, and the possibility degree of character pixel points:

[0174]

[0175] Among them, n represents the growth satisfaction degree; m represents the difference between the horizontal distance D between the edge pixel points and the standard horizontal distance, and j represents the possibility degree of character pixel points.

[0176] Among them, the larger n is, the more it is considered that the horizontal distance between the edge pixel points meets the reference character standard. The growth satisfaction degree can be compared with the preset growth satisfaction degree threshold. For example: the preset growth satisfaction degree threshold is 0.8. If n is greater than or equal to 0.8, the region growth stops.

[0177] In this embodiment, by extracting edge pixel points, establishing edge vectors, calculating the difference in angle changes, and evaluating the likelihood of character pixel points, it is possible to ensure that region growing proceeds along the stroke direction of the character, improving the accuracy of character recognition. By setting a threshold and comprehensively evaluating the likelihood of character pixel points and the horizontal distance, background and noise points can be effectively filtered out, avoiding mis-adding irrelevant pixel points to the target region and reducing the possibility of misrecognition.

[0178] In an embodiment of the present application, the preset verification condition includes a preset threshold for the degree of coincidence of character inflection points. Verifying the target character based on the matching character, and when the result of the verification meets the preset verification condition, taking the target character as the result of license plate auxiliary recognition includes:

[0179] Comparing the inflection points of the matching character with the inflection points of the target character to obtain the degree of coincidence of character inflection points;

[0180] When the degree of coincidence of character inflection points is greater than or equal to the preset threshold for the degree of coincidence of character inflection points, determining that the target character is the result of license plate auxiliary recognition.

[0181] Among them, an inflection point refers to a point where the direction changes significantly in the character contour, usually a turning point or a cusp of the character. By methods such as edge detection and curve fitting, the inflection points on the character contour are extracted. Inflection points are important feature points of characters and are used to distinguish different character shapes.

[0182] Among them, the degree of coincidence of inflection points refers to the similarity between the inflection points of the matching character and the inflection points of the target character. By calculating parameters such as the distance and angle between the inflection points of the matching character and the target character, its degree of coincidence is evaluated. Through the degree of coincidence of inflection points, the consistency between the target character and the matching character is verified to ensure the accuracy of the recognition result.

[0183] Exemplarily, the preset threshold for the degree of coincidence of character inflection points is 0.87. If the degree of coincidence of character inflection points is greater than or equal to 0.87, then the target character is determined to be the result of license plate auxiliary recognition.

[0184] In this embodiment, by extracting and comparing inflection points, the consistency between the target character and the matching character can be accurately verified, reducing the possibility of misrecognition and improving the accuracy of character recognition. The calculation method of the degree of coincidence of inflection points can effectively filter out characters with inconsistent shapes, avoiding misidentifying noise or background information as valid characters and reducing misrecognition.

[0185] Figure 4 It is a schematic structural diagram of a license plate auxiliary recognition system for an intelligent checkpoint provided by an embodiment of the present invention. This system can be applied to Figure 1The implementation environment shown. This system can also be applied to other exemplary implementation environments and be specifically configured in other devices. This embodiment does not limit the implementation environment applicable to this system.

[0186] As Figure 4 shown, the exemplary license plate auxiliary recognition system of the intelligent checkpoint includes:

[0187] An image acquisition module 401, configured to acquire a front image of a vehicle collected by the intelligent checkpoint;

[0188] A contour detection module 402, configured to perform contour detection on the front image of the vehicle, and obtain a license plate image according to the result of the contour detection, where the license plate image includes a plurality of characters to be processed;

[0189] A region growing module 403, configured to, for any one of the characters to be processed, match the character to be processed with a preset character library to obtain a matching character, and perform region growing on the periphery of the character to be processed according to the shape feature of the matching character to obtain a target character;

[0190] A character verification module 404, configured to verify the target character according to the matching character, and when the result of the verification meets a preset verification condition, use the target character as the result of license plate auxiliary recognition, and perform traffic control according to the result of the license plate auxiliary recognition.

[0191] In the exemplary license plate auxiliary recognition system of the intelligent checkpoint, the character to be processed is matched with a preset character library to obtain a matching character; then, region growing is performed on the surrounding area according to the shape feature of the matching character to obtain a complete target character. This license plate auxiliary recognition method can cope with situations such as occlusion and damage of license plate characters, and improve the accuracy of character recognition. Through accurate license plate auxiliary recognition results, the intelligent checkpoint can accurately identify illegal vehicles, automatically record them and trigger corresponding traffic control measures, such as fines and points deduction; the intelligent checkpoint can, through accurate license plate auxiliary recognition results, analyze traffic data in real time and accurately, provide timely and accurate information support for the traffic management department, help the traffic management department make decisions quickly, optimize traffic flow, and improve road traffic efficiency.

[0192] It should be noted that the license plate auxiliary recognition system of the intelligent checkpoint provided in the above embodiments and the license plate auxiliary recognition method of the intelligent checkpoint provided in the above embodiments belong to the same concept. The specific ways in which each module and unit perform operations have been described in detail in the method embodiments and will not be repeated here. In practical applications, the license plate auxiliary recognition system of the intelligent checkpoint provided in the above embodiments can allocate the above functions to different functional modules as needed, that is, divide the internal structure of the system into different functional modules to complete all or part of the functions described above, and this will not be limited here either.

[0193] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0194] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments.

Claims

1. A smart card slot auxiliary license plate recognition method, characterized in that: The method comprises: Obtain the front image of the vehicle captured by the smart camera; Performing contour detection on the front image of the vehicle, and obtaining a license plate image according to the result of the contour detection, wherein the license plate image includes a plurality of characters to be processed; For any of the characters to be processed, the character to be processed is matched with a preset character library to obtain a matching character, and region growth is performed around the character to be processed according to shape features of the matching character to obtain a target character; According to the matching characters, the target characters are verified, and when the verification result meets the preset verification condition, the target characters are used as the result of the license plate auxiliary recognition, and traffic control is performed according to the result of the license plate auxiliary recognition; The step of matching the character to be processed with a preset character library to obtain a matching character includes: Determining a preset character library corresponding to the character to be processed according to the position of the character to be processed, and calculating the matching degree between the character to be processed and a plurality of reference characters in the preset character library; Determining the matching character from the plurality of reference characters according to the matching degree; The calculating the matching degree between the character to be processed and a plurality of reference characters in the preset character library comprises: Performing corner point detection on the character to be processed to obtain a first corner point set corresponding to the character to be processed; Performing corner point detection on each of the reference characters to obtain a second corner point set corresponding to each of the reference characters; Mapping the positions of corner points in the first corner point set to the second corner point set, and determining corner points in the second corner point set that meet a preset distance threshold around the mapped positions; The matching degree between the character to be processed and each of the reference characters is obtained according to the number of corner points in the second corner point set that meet a preset distance threshold around the mapped position.

2. The license plate auxiliary recognition method of the smart card port as claimed in claim 1, characterized in that: After obtaining the license plate image, the method includes: When the license plate image is in color, grayscale processing is performed on the license plate image to obtain a grayscale image corresponding to the license plate image; Performing dilation processing on the grayscale image by using a preset structural element to obtain a grayscale image after dilation processing; Performing an erosion process on the grayscale image after the dilation process by using the preset structural element to obtain an eroded grayscale image; The grayscale image after the corrosion process is subjected to threshold segmentation process to obtain a binary license plate image, and the binary license plate image is used to match with the preset character library.

3. The license plate auxiliary recognition method of the smart card port as claimed in claim 1, characterized in that: The step of performing region growing around the character to be processed according to the shape feature of the matching character comprises: When a first corner point in the first corner point set is positionally mapped to the second corner point set, if there are corner points around that meet a preset distance threshold, the first corner point is used as a seed point; Setting a gray value difference threshold when filling pixels around the seed point, and setting a region growth direction according to a stroke feature corresponding to the position of the seed point in the matching character; For any of the seed points, region growing is performed around the seed point according to the gray value difference threshold and the region growing direction.

4. The license plate auxiliary recognition method of the smart card port as claimed in claim 3, characterized in that: The performing region growing around the seed point according to the gray value difference threshold and the region growing direction includes: According to the gray value difference threshold, a plurality of pixel points are determined from around the seed point to obtain an area to be filled; Performing grayscale filling on the pixel points in the area to be filled, and connecting the seed point with the pixel points after grayscale filling to obtain a growth vector; Determining the growth possibility according to the angle between the growth direction of the region and the growth vector; The growth possibility is compared with a preset growth possibility threshold, and the region growing process is controlled according to the comparison result.

5. The license plate auxiliary recognition method of the smart card port as claimed in claim 4, characterized in that: The controlling of the region growing process according to the comparison result comprises: When the comparison result shows that the growth possibility is greater than or equal to the preset growth possibility threshold, the pixel point after grayscale filling is used as a new seed point, and regional growth is performed around the new seed point according to the grayscale value difference threshold; When the comparison result is that the growth possibility is less than the preset growth possibility threshold, the region growing around the seed point is stopped.

6. The license plate auxiliary recognition method of the smart card port as claimed in claim 5, characterized in that: After performing region growing around the new seed point, the method further includes: Based on a new character region obtained by performing region growth around the new seed point, a plurality of edge vectors are established along the region growth direction for edge pixel points of the new character region, and the angle change difference between each two adjacent edge vectors is calculated; According to the angle change difference, the probability of the character pixel point corresponding to the edge pixel point is obtained; Determine the lateral distance between the spaced edge pixel points along the region growth direction; The process of the region growth is controlled by combining the possible degree of the character pixel points and the lateral distance.

7. The license plate auxiliary recognition method of the smart card port as claimed in claim 1, characterized in that: The preset verification condition includes a preset character inflection point matching degree threshold, and the target character is verified according to the matching character, and when the result of the verification meets the preset verification condition, the target character is used as the result of the license plate auxiliary recognition, including: Comparing the inflection point of the matching character with the inflection point of the target character to obtain the degree of coincidence of the inflection points of the characters; When the degree of matching of the inflection points of the characters is greater than or equal to the preset threshold value of the degree of matching of the inflection points of the characters, it is determined that the target character is the result of the auxiliary license plate recognition.

8. A smart card slot license plate auxiliary recognition system, characterized in that: The system comprises: An image acquisition module, used to acquire the front image of the vehicle captured by the smart camera; A contour detection module, used to perform contour detection on the front image of the vehicle, and obtain a license plate image according to the result of the contour detection, wherein the license plate image includes a plurality of characters to be processed; A region growing module, for matching any of the characters to be processed with a preset character library to obtain a matching character, and performing region growing around the character to be processed according to shape features of the matching character to obtain a target character; The step of matching the character to be processed with a preset character library to obtain a matching character includes: Determining a preset character library corresponding to the character to be processed according to the position of the character to be processed, and calculating the matching degree between the character to be processed and a plurality of reference characters in the preset character library; Determining the matching character from the plurality of reference characters according to the matching degree; The calculating the matching degree between the character to be processed and a plurality of reference characters in the preset character library comprises: Performing corner point detection on the character to be processed to obtain a first corner point set corresponding to the character to be processed; Performing corner point detection on each of the reference characters to obtain a second corner point set corresponding to each of the reference characters; Mapping the positions of corner points in the first corner point set to the second corner point set, and determining corner points in the second corner point set that meet a preset distance threshold around the mapped positions; Obtaining a degree of matching between the character to be processed and each of the reference characters according to the number of corner points in the second corner point set that meet a preset distance threshold around the mapped position; The character verification module is used to verify the target character according to the matching character, and when the verification result meets the preset verification condition, the target character is used as the result of the license plate auxiliary recognition, and traffic control is performed according to the result of the license plate auxiliary recognition.

Images