A large-angle license plate recognition method, device, equipment and storage medium

Abstract

The application discloses a large-angle license plate recognition method, device and equipment and a storage medium, and the large-angle license plate recognition method comprises the following steps: acquiring image information, acquiring a vehicle head identification box and a license plate identification box on the image information based on a deep learning model; acquiring a vehicle head direction and a skew transformation angle according to the vehicle head identification box and the license plate identification box; acquiring a skew transformation matrix according to the vehicle head direction and the skew transformation angle; performing skew transformation on the image information according to the skew transformation matrix, and acquiring transformed image information; and recognizing a license plate number according to the transformed image information. The application solves the problems of low precision and low robustness of the existing license plate recognition mode, and improves the recognition accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image recognition, and in particular to a large-angle license plate recognition method, device, equipment and storage medium. BACKGROUND

[0002] The license plate recognition system is an application of computer vision image recognition technology in vehicle photographing recognition. The license plate recognition technology requires that the license plate of a moving vehicle can be extracted and recognized from a complex background. Through license plate extraction, image preprocessing, feature extraction, license plate character recognition and other technologies, the license number, color and other information of the vehicle can be recognized.

[0003] The existing method mainly uses four-point detection and affine transformation to recognize the large-angle license plate. This method relies on accurate four-point positioning of the license plate and does not have strong robustness for license plates of different sizes in different scenes. In addition, since the shape of the license plate is not the same at a large angle of view, it is a parallelogram with an indefinite angle, the four-point positioning method has limitations in precision.

[0004] Therefore, there is a need for a new technical solution. SUMMARY

[0005] Therefore, the embodiments of the present application provide a large-angle license plate recognition method, device, equipment and storage medium to at least solve the problems of low precision and low robustness of the existing license plate recognition method.

[0006] The embodiments of the present application provide the following technical solutions:

[0007] The large-angle license plate recognition method provided by the embodiments of the present application comprises:

[0008] Obtaining image information, and obtaining a vehicle head bounding box and a license plate bounding box based on a deep learning model on the image information;

[0009] Obtaining a vehicle head direction and a skew transformation angle according to the vehicle head bounding box and the license plate bounding box;

[0010] Obtaining a skew transformation matrix according to the vehicle head direction and the skew transformation angle;

[0011] Performing skew transformation on the image information according to the skew transformation matrix, and obtaining transformed image information;

[0012] Recognizing a license plate number according to the transformed image information.

[0013] Further, the obtaining of the vehicle head direction based on the vehicle head bounding box and the license plate bounding box comprises:

[0014] Obtaining a vehicle head vertex coordinate of the vehicle head bounding box and a license plate vertex coordinate of the license plate bounding box;

[0015] obtaining the head direction based on the head vertex coordinates and the license plate vertex coordinates.

[0016] Further, when the head vertex coordinates include first head vertex coordinates (x1, y1) and second head vertex coordinates (x2, y2), and the license plate vertex coordinates include first license plate vertex coordinates (x3, y3) and second license plate vertex coordinates (x4, y4), the calculation formula of the head direction is as follows:

[0017]

[0018] Wherein, From is used to represent the head direction, in the case of From equal to 1, the head is to the left; in the case of From equal to -1, the head is to the right.

[0019] Further, the obtaining the skew transformation angle includes:

[0020] obtaining the license plate width, and obtaining the skew transformation angle based on the license plate vertex coordinates and the license plate width.

[0021] Further, the obtaining the license plate width includes:

[0022] obtaining the license plate width based on a Hough line transformation algorithm.

[0023] Further, when the license plate vertex coordinates include third license plate vertex coordinates (x3, y3) and fourth license plate vertex coordinates (x4, y4), the formula for obtaining the skew transformation angle is as follows:

[0024]

[0025] Wherein, α is the skew transformation angle, and h is the license plate width.

[0026] Further, when the pre-transformation coordinates of a point in the image information are (x, y), and the post-transformation coordinates are (x', y'), the relationship between the pre-transformation coordinates and the post-transformation coordinates is as follows:

[0027] x'x'=x;

[0028] y'=y-From*x*tanα;

[0029] Wherein, From is used to represent the head direction, and α is the skew transformation angle.

[0030] The large-angle license plate recognition device provided by the embodiment of the application comprises:

[0031] An acquisition unit is configured to acquire image information, acquire a vehicle head bounding box and a license plate bounding box based on a deep learning model, acquire a vehicle head orientation and a skew transformation angle according to the vehicle head bounding box and the license plate bounding box, and acquire a skew transformation matrix according to the vehicle head orientation and the skew transformation angle;

[0032] A transformation unit is configured to perform skew transformation on the image information according to the skew transformation matrix, and acquire transformed image information;

[0033] An identification unit is configured to identify a license plate number according to the transformed image information.

[0034] A computer device of an embodiment of the present application includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the large-angle license plate recognition method of any one of the above when executing the computer program.

[0035] A computer-readable storage medium of an embodiment of the present application has a computer program stored thereon, and the computer program is executed by a processor to implement the large-angle license plate recognition method of any one of the above.

[0036] Compared with the prior art, the above at least one technical solution adopted by the embodiment of the present application can achieve at least the following beneficial effects:

[0037] The large-angle license plate recognition method of the present application acquires image information, acquires a vehicle head bounding box and a license plate bounding box based on a deep learning model, acquires a vehicle head orientation and a skew transformation angle according to the vehicle head bounding box and the license plate bounding box, acquires a skew transformation matrix according to the vehicle head orientation and the skew transformation angle, performs skew transformation on the image information according to the skew transformation matrix, and acquires transformed image information, and identifies a license plate number according to the transformed image information, thereby solving the problems of low precision and low robustness of the existing license plate recognition method, and improving the recognition accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0039] Fig. 1 Flowchart (I) of the large-angle license plate recognition method of the embodiment of the present application;

[0040] Fig. 2 Flowchart (II) of the large-angle license plate recognition method of the embodiment of the present application;

[0041] Fig. 3 The specific implementation structure diagram of the large-angle license plate recognition method of the embodiment of the present application;

[0042] Fig. 4 The structure diagram of the vehicle head identification frame and the license plate identification frame of the embodiment of the present application;

[0043] Fig. 5 The simulation diagram (I) of the embodiment of the present application;

[0044] Fig. 6 The simulation diagram (II) of the embodiment of the present application;

[0045] Fig. 7 The structure block diagram of the large-angle license plate recognition device of the embodiment of the present application. DETAILED DESCRIPTION

[0046] The embodiments of the present application will be described in detail below with reference to the drawings.

[0047] The above embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application. The present application can also be implemented or applied by using other different specific embodiments, and each detail in the specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0048] It should be noted that the various aspects of the embodiments described below are within the scope of the appended claims. It should be apparent that the aspects described herein can be embodied in a wide variety of forms and that any specific structure and / or function described herein is merely illustrative. Based on the teachings herein one skilled in the art should appreciate that an aspect described herein can be implemented independently of any other aspects and that an aspect described herein can be implemented both as any number of software and / or hardware structures and as any number of combinations of software and / or hardware structures. For example, an apparatus can be implemented using any number and combination of the aspects described herein. Additionally, the features described herein can be implemented using software and / or hardware structures that are not expressly described herein. For example, a device can be implemented using any number and combination of the aspects described herein.

[0049] It is also necessary to point out that the drawings provided in the following embodiments only illustrate the basic concepts of the present application in a schematic manner, and only show the components related to the present application in the drawings, not drawn according to the number, shape and size of the components in actual implementation, the actual implementation of each component can be a random change in type, number and proportion, and the component layout type can also be more complex.

[0050] In addition, in the following description, specific details are provided in order to facilitate a thorough understanding of the examples. However, one skilled in the art will understand that the practice can be practiced without these specific details.

[0051] In some operation scenarios, the camera for shooting the license plate cannot be installed in a suitable position. For example, in port road testing, "lifting the box" is an important production link, various types of container trucks drive to the operation site to complete the lifting and placing operations. Before the operation, it is necessary to confirm the license plate information of the current container truck to determine the operation situation. The camera of the road test starts shooting after the container truck drives into the operation area, and sends the image to the recognition program at a fixed frequency per second. Due to the installation angle of the camera, the license plate photo taken by the road test often has a large angle, and due to the non-fixed operation site, the proportion and position of the license plate in the whole picture are not fixed, which increases the difficulty of recognizing the large-angle license plate.

[0052] Therefore, the present application uses artificial intelligence and computer vision recognition technology to solve the problem of large-angle license plate (license plate inclination angle greater than 45 degrees) recognition through the method of vehicle head positioning-license plate positioning-affine transformation-text recognition, to improve the accuracy of license plate recognition.

[0053] Based on this, the embodiment of the present application proposes a processing scheme: the present application identifies the vehicle head and obtains the orientation of the vehicle head, which helps to obtain the plane where the license plate is located and the distortion mode of the license plate, so as to perform the shearing transformation on the image according to the plane and the distortion mode of the license plate to correct the image, thereby improving the accuracy of recognizing the license plate information. At the same time, during the license plate detection and recognition process, understanding the orientation of the vehicle head can also more accurately analyze the license plate area and the character arrangement direction, which helps to exclude irrelevant background areas, reduces the risk of misrecognition, and improves the recognition accuracy.

[0054] The technical solutions provided by the embodiments of the present application are described below in conjunction with the drawings.

[0055] As Figs. 1-2 shown, the embodiment of the present application provides a large-angle license plate recognition method, which comprises:

[0056] Step S102, obtaining image information, obtaining a vehicle head bounding box and a license plate bounding box on the image information based on a deep learning model;

[0057] Step S104, obtaining the vehicle head orientation and skew transformation angle according to the vehicle head identification frame and the license plate identification frame;

[0058] Step S106, obtaining the skew transformation matrix according to the vehicle head orientation and the skew transformation angle;

[0059] Step S108, performing skew transformation on the image information according to the skew transformation matrix, and obtaining transformed image information;

[0060] Step S110, recognizing the license plate number according to the transformed image information.

[0061] In step S102, the image information is obtained by a camera, and at least includes vehicle head information and license plate information.

[0062] The deep learning model is a model based on deep learning, and is trained by deep learning using images labeled with vehicle heads or license plates.

[0063] Specifically, the target detection of the vehicle head is performed based on the yolo-v5 method based on deep learning, and the deep learning is trained based on images labeled with vehicle heads or license plates. The backbone network of the yolo-v5 model is mainly composed of a Focus module, a BottleneckCSP network. First, an image with a size of 1920*1440*3 is input, and the image is sliced by the Focus module. Then, the integrated image is subjected to feature extraction by a Conv module with a size of 3 and a step of 2. Subsequently, a BottleneckCSP2 module is used to reduce the model parameter quantity. The 80*80*512 size feature map is up-sampled and subjected to convolution, and finally three feature maps with sizes of 80*80*255, 40*40*255 and 20*20*255 are obtained. Finally, candidate boxes are generated on the three different scale feature maps, and the target boxes are screened by using a weighted non-maximum value method (in the present application, it is divided into "vehicle head" or "non-vehicle head").

[0064] The vehicle head identification frame is used to identify the vehicle head in the image information, and the license plate identification frame is used to identify the license plate in the image information.

[0065] In step S104, accurately determining the vehicle head orientation helps to understand the plane where the license plate is located, and corrects the image according to the plane to correct the license plate, so as to more accurately analyze the license plate area and license plate characters, and improve the recognition rate.

[0066] Wherein, by understanding the head orientation, it is also helpful to exclude some irrelevant background areas, reduce the risk of misidentification, and help process complex scenes or multiple vehicle models, provide more stable license plate detection for different types of vehicles, thereby helping to improve the accuracy and stability of license plate detection and recognition.

[0067] Wherein, the head orientation is acquired to help understand the position and direction of the vehicle, thereby providing information about the possible perspective angle of the license plate, and then understanding the position and angle of the vehicle relative to the camera, to calculate or predict the degree and direction of deformation of the vehicle using this information.

[0068] Wherein, the method of acquiring the head orientation includes methods such as identifying the head orientation based on features, identifying the head orientation based on deep learning, identifying the head orientation based on optical flow, and identifying the head orientation based on stereo vision.

[0069] In some embodiments of the present application, as shown in Fig. 2 The head abstract based on the head bounding box and the license plate bounding box is acquired by the following steps:

[0070] Step S104a, acquiring the head vertex coordinates of the head bounding box and the license plate vertex coordinates of the license plate bounding box;

[0071] Step S104b, acquiring the head orientation based on the head vertex coordinates and the license plate vertex coordinates.

[0072] In step S104a, the head vertex coordinates can include the top-left corner vertex coordinates of the head bounding box and the bottom-right corner vertex coordinates of the head bounding box, and the license plate vertex coordinates include the top-left corner vertex coordinates of the license plate bounding box and the bottom-right corner vertex coordinates of the license plate bounding box.

[0073] In a preferred embodiment, as shown in Figs. 3-4 When the head vertex coordinates include the first head vertex coordinates (x1, y1) and the second head vertex coordinates (x2, y2), and the license plate vertex coordinates include the first license plate vertex coordinates (x3, y3) and the second license plate vertex coordinates (x4, y4), the calculation formula of the head orientation is as follows:

[0074]

[0075] Wherein, From is used to represent the head orientation, in the case of From equal to 1, the head is to the left; in the case of From equal to -1, the head is to the right.

[0076] Wherein, the first vehicle head vertex coordinate (x1, y1) can be the top left corner vertex coordinate of the vehicle head identification frame, and the second vehicle head vertex coordinate (x2, y2) can be the bottom right corner vertex coordinate of the vehicle head identification frame; the first license plate vertex coordinate (x3, y3) is the top left corner vertex coordinate of the license plate identification frame, and the second license plate vertex coordinate (x4, y4) is the bottom right corner vertex coordinate of the license plate identification frame.

[0077] In some embodiments, according to the position change of the camera, the first vehicle head vertex coordinate (x1, y1) can be the top right corner vertex coordinate of the vehicle head identification frame, and the second vehicle head vertex coordinate (x2, y2) can be the bottom left corner vertex coordinate of the vehicle head identification frame; the first license plate vertex coordinate (x3, y3) is the top right corner vertex coordinate of the license plate identification frame, and the second license plate vertex coordinate (x4, y4) is the bottom left corner vertex coordinate of the license plate identification frame.

[0078] In some embodiments of the present application, the step of obtaining the skew transformation angle comprises:

[0079] Step S104c, obtaining the license plate width, and obtaining the skew transformation angle based on the license plate vertex coordinate and the license plate width.

[0080] Wherein, the license plate width can be obtained based on the Hough line transformation algorithm.

[0081] Specifically, as shown in the figure, Figs. 5-6 When the license plate width h is obtained by using the straight line detection Hough_line algorithm, for a given point (x0, y0), all straight lines passing through it can be drawn in the rectangular coordinate system (θ is the x-axis and r is the y-axis). Finally, we will get a sinusoidal curve.

[0082] If the curves obtained after the above operation of two different points intersect on the plane, it means that there is a common (θ, rθ) between the two different points, that is, a common straight line. The more curves intersect at a point, the more points the straight line represented by the intersection point consists of, and the greater the weight. The Hough line transformation algorithm tracks the intersection points between the curves corresponding to each point in the image. If the number of curves intersecting at a point exceeds the threshold, then the parameters represented by the intersection point are considered to be a straight line in the original image.

[0083] More specifically: create a two-dimensional array to save the values of the two parameters, use r to represent the row and θ to represent the column; the size of the array depends on the required accuracy. In this scheme, the accuracy is set to 0.5°, that is, 360 columns of arrays are opened to record. Traverse r and θ, and calculate the line segment that meets the license plate slope range, and finally obtain the license plate width h.

[0084] In some embodiments, after obtaining the license plate width, when the license plate vertex coordinates include a third license plate vertex coordinate (x3, y3) and a fourth license plate vertex coordinate (x4, y4), the formula for obtaining the skew transformation angle is as follows:

[0085]

[0086] wherein a is the skew transformation angle, and h is the license plate width.

[0087] In step S106, after obtaining the vehicle heading and the skew transformation angle, matrix decomposition can be performed based on the python getPerspectiveTransform to calculate the perspective matrix X that satisfies AX = B, wherein A is the detected license plate four-point coordinate matrix, and B is the specific position to be stretched into a rectangle.

[0088] In some specific embodiments of the present application, when the pre-transformation coordinates of a point in the image information are (x, y) and the post-transformation coordinates are (x', y'), the relationship between the pre-transformation coordinates and the post-transformation coordinates is as follows:

[0089] x' = x;

[0090] y' = y - From * tan a;

[0091] wherein From represents the vehicle heading, and a is the skew transformation angle.

[0092] wherein the post-transformation coordinates are obtained through the pre-transformation coordinates, so that skew transformation can be performed on all points of the image to obtain the skew transformation matrix.

[0093] In step S108, the skew transformation is used to stretch the image information into a rectangle, so as to facilitate subsequent recognition of the license plate number.

[0094] Specifically, the extracted license plate picture can be mapped to the corresponding position of the rectangle by using the warpPerspective function of python, so as to facilitate subsequent text recognition detection.

[0095] In step S110, after obtaining the standard size of the rectangular image, the picture is input into the text recognition system.

[0096] Specifically, the base model used in step S110 is a ResNet (Residual Network) to extract basic features of the text. The output obtained is a tensor with a size of (8, 512, 8, 32), where 8 represents a batch size, 512 represents the number of feature maps, and 8*32 represents the size of the feature map. Subsequently, the feature map extracted by the Resnet is added to the Unet network (U-shaped network) and then sent to a fully connected layer after a position attention mechanism in which the information of each letter position is fixed, to obtain the final classification of the character.

[0097] The application does not need to consider the distance of the camera from the vehicle in the large-angle license plate recognition task, has high robustness, high recognition accuracy of large-angle license plates, and has beneficial effects on the promotion of large-angle license plate recognition tasks in different scenes.

[0098] As Fig. 7 shown, the embodiment of the application also provides a large-angle license plate recognition device, which comprises an acquisition unit 10, a transformation unit 20 and an identification unit 30. The acquisition unit 10 is configured to acquire image information, acquire a vehicle head identification frame and a license plate identification frame based on a deep learning model, acquire a vehicle head orientation and a skew transformation angle according to the vehicle head identification frame and the license plate identification frame, and acquire a skew transformation matrix according to the vehicle head orientation and the skew transformation angle. The transformation unit 20 is configured to perform skew transformation on the image information according to the skew transformation matrix and acquire transformed image information. The identification unit 30 is configured to identify a license plate number according to the transformed image information.

[0099] In some embodiments, the acquisition unit 10 is further configured to acquire vehicle head vertex coordinates of the vehicle head identification frame and license plate vertex coordinates of the license plate identification frame, and acquire the vehicle head orientation based on the vehicle head vertex coordinates and the license plate vertex coordinates.

[0100] In some embodiments, the acquisition unit 10 is further configured to acquire a license plate width, and acquire the skew transformation angle based on the license plate vertex coordinates and the license plate width.

[0101] In some embodiments, the acquisition unit 10 is further configured to acquire the license plate width based on a Hough line transformation algorithm.

[0102] The embodiment of the application also provides a computer device. The computer device can include one or more processors and a memory.

[0103] The processor can be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and can control other components in the computer device to perform desired functions.

[0104] The memory can include one or more computer programs, which can include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory, for example, can include random access memory (RAM), cache memory, and / or the like. The non-volatile memory, for example, can include read only memory (ROM), hard disk, flash memory, and / or the like. One or more computer programs can be stored on the computer-readable storage media, and the processor can execute the program instructions to implement the above-mentioned large-angle license plate recognition method and / or other desired functions.

[0105] In addition, the electronic device can also include any other appropriate components, such as a bus, an input / output interface, and the like, according to specific application circumstances.

[0106] In addition to the above-mentioned method and device, the embodiments of the present disclosure can also be a computer program product, which includes computer program instructions that, when executed by a processor, cause the processor to perform the steps of the large-angle license plate recognition method described above in the specification.

[0107] The computer program product can be written in any combination of one or more programming languages, including an object-oriented programming language, such as Java, C++, and the like, and conventional procedural programming languages, such as the "C" programming language, or the like. The program code can be executed entirely on the user computing device, partially on the user device, as a stand-alone software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0108] In addition, the embodiments of the present disclosure can also be a computer-readable storage medium, which stores computer program instructions, and the computer program instructions, when executed by a processor, cause the processor to perform the steps of the large-angle license plate recognition method described above in the specification.

[0109] The computer readable storage medium can be any combination of one or more computer readable medium(s). The computer readable medium can be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include the following: an electrical connection having one or more wires, a portable disc, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0110] In this specification, the same or similar parts among various embodiments are referred to each other, and each embodiment focuses on the difference from other embodiments. Especially, for the product embodiment described later, since it is corresponding to the method, the description is simple, and the relevant part refers to the part of the system embodiment.

[0111] The above merely illustrates the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of the change or replacement within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A large angle license plate recognition method, characterized by, The method comprises the following steps: obtaining image information through a camera, and obtaining a vehicle head identification frame and a license plate identification frame on the image information based on a deep learning model, wherein the image information at least includes vehicle head information and license plate information; obtaining a vehicle head direction and a skew transformation angle according to the vehicle head identification frame and the license plate identification frame, wherein the skew transformation angle is an angle between the license plate and the horizontal direction; obtaining a skew transformation matrix according to the vehicle head direction and the skew transformation angle; skew transforming the image information according to the skew transformation matrix, and obtaining transformed image information; recognizing a license plate number according to the transformed image information; wherein obtaining the vehicle head direction according to the vehicle head identification frame and the license plate identification frame comprises: obtaining a vehicle head vertex coordinate of the vehicle head identification frame and a license plate vertex coordinate of the license plate identification frame; obtaining the vehicle head direction based on the vehicle head vertex coordinate and the license plate vertex coordinate; the method for obtaining the skew transformation angle comprises obtaining a license plate width based on a Hough line transformation algorithm, and obtaining the skew transformation angle based on the license plate vertex coordinate and the license plate width; the method for obtaining the vehicle head direction is also used to help exclude irrelevant background areas, reduce the risk of misrecognition, and process complex scenes or multiple vehicle types; and is also used to help understand the position and direction of the vehicle, provide possible perspective angle information of the license plate, and then understand the position and angle of the vehicle relative to the camera, so as to calculate or predict the deformation degree and direction of the vehicle.

2. The large angle license plate recognition method according to claim 1, wherein, when the vehicle head vertex coordinate comprises a first vehicle head vertex coordinate (x1, y1) and a second vehicle head vertex coordinate (x2, y2), and the license plate vertex coordinate comprises a first license plate vertex coordinate (x3, y3) and a second license plate vertex coordinate (x4, y4), the calculation formula of the vehicle head direction is as follows: ； wherein From is used to represent the vehicle head direction, in the case that From is equal to 1, the vehicle head is to the left; and in the case that From is equal to -1, the vehicle head is to the right.

3. The large angle license plate recognition method of claim 1, wherein, when the license plate vertex coordinate comprises a third license plate vertex coordinate (x3, y3) and a fourth license plate vertex coordinate (x4, y4), the formula for obtaining the skew transformation angle is as follows: ； wherein, is the skew angle and h is the license plate width.

4. The large angle license plate recognition method of claim 1, wherein, when the pre-transformation coordinate of a point in the image information is (x, y), and the post-transformation coordinate is (x', y'), the relationship between the pre-transformation coordinate and the post-transformation coordinate is as follows: x'=x; ； Wherein, From is used to indicate the vehicle head direction, is the skew angle.

5. A wide angle license plate recognition device, characterized by, The method comprises the following steps: an obtaining unit is used to obtain image information through a camera, obtain a vehicle head identification frame and a license plate identification frame based on a deep learning model, obtain a vehicle head direction and a skew transformation angle according to the vehicle head identification frame and the license plate identification frame, and obtain a skew transformation matrix according to the vehicle head direction and the skew transformation angle, wherein the skew transformation angle is an angle between the license plate and the horizontal direction, and the method for obtaining the skew transformation angle comprises obtaining a license plate width; a transforming unit is used to skew transform the image information according to the skew transformation matrix, and obtain transformed image information; an identifying unit is used to recognize a license plate number according to the transformed image information. The obtaining unit is further configured to obtain head vertex coordinates of the head identification frame and license plate vertex coordinates of the license plate identification frame, and obtain the head direction based on the head vertex coordinates and the license plate vertex coordinates. The obtaining unit is further configured to obtain the skewing and shearing transformation angle based on the license plate vertex coordinates and the license plate width, and the image information at least includes head information and license plate information. The obtaining head direction is further configured to help exclude irrelevant background areas, reduce the risk of misrecognition, and process complex scenes or multiple vehicle types; and is further configured to help understand the position and direction of the vehicle, provide possible perspective angle information about the license plate, and then understand the position and angle of the vehicle relative to the camera to calculate or predict the deformation degree and direction of the vehicle.

6. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the large-angle license plate recognition method according to any one of claims 1 to 4.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the large-angle license plate recognition method according to any one of claims 1 to 4.

Images