License plate number identification method and related device

Abstract

The embodiment of the invention provides a license plate number identification method and a related device. The method comprises the following steps: acquiring a license plate image to be identified; a license plate part in the license plate image is positioned by adopting a YOLOv5-Mobi leNet model to obtain a license plate area, the YOLOv5-Mobi leNet model is obtained by combining YOLOv5s and Mobi leNet, the YOLOv5s is used for rapidly detecting the rough position of the license plate, and the Mobi leNet accurately positions the license plate area through a lightweight convolutional neural network; correcting the inclined license plate area by using a spatial transformation network to obtain a corrected license plate area; and inputting the corrected license plate area into the trained CRNN-LPR character recognition network, performing license plate character recognition, extracting a license plate number, and outputting a license plate number recognition result. According to the embodiment of the invention, the accuracy of license plate number detection and recognition can be improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical fields of computer vision and artificial intelligence, and particularly to a license plate number recognition method and related device. Background Art

[0002] A license plate recognition system is an automated system based on computer vision and artificial intelligence technologies, used to recognize the license plate information of vehicles. Its main function is to capture the license plate of a vehicle through an image or video stream and extract the character information therein to achieve the identification and tracking of the vehicle. License plate recognition systems are usually applied to scenarios such as traffic management, parking lot systems, electronic toll collection, and highway monitoring.

[0003] Currently, the license plate detection and recognition technology adopted by license plate recognition systems has technical defects such as inaccurate positioning, low recognition efficiency, and poor real-time performance for the recognition of double-layer license plates. Therefore, how to effectively improve the accuracy and real-time performance of license plate recognition has become an urgent technical problem to be solved. Summary of the Invention

[0004] The embodiments of the present invention provide a license plate number recognition method and related device, which can effectively improve the accuracy and real-time performance of license plate recognition.

[0005] In a first aspect, the embodiments of the present invention provide a license plate number recognition method, including:

[0006] Obtain a license plate image to be recognized;

[0007] Use the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area, where the YOLOv5-MobileNet model is obtained by combining YOLOv5s and MobileNet, the YOLOv5s is used to quickly detect the rough position of the license plate, and the MobileNet accurately locates the license plate area through a lightweight convolutional neural network;

[0008] Use a spatial transformation network to correct the tilted license plate area to obtain the corrected license plate area;

[0009] Input the corrected license plate area into a trained CRNN-LPR character recognition network to perform license plate character recognition and extract the license plate number, and output the license plate number recognition result.

[0010] In some embodiments, the step of using the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area includes:

[0011] Replace the original backbone network in YOLOv5 with MobileNetV3 to locate the license plate part in the license plate image and obtain the license plate area.

[0012] In some embodiments, after using the YOLOv5-MobileNet model to locate the license plate part in the license plate image and obtain the license plate area, the following steps are further included:

[0013] Use a feature pyramid network to fuse multi-layer features of the license plate area and output multi-scale fusion features.

[0014] In some embodiments, the CRNN-LPR character recognition network includes an improved CRNN, LPRNet, and LocNet. The CRNN-LPR character recognition network uses a fully convolutional layer network and adds a CTC loss function. The fully convolutional network replaces the Bi-LSTM module in the CRNN, and the probability distribution of license plate characters is output through the fully convolutional layer. The LocNet is used to improve the accuracy of license plate character recognition.

[0015] In some embodiments, the YOLOv5-MobileNet model is deployed on an embedded device.

[0016] In some embodiments, the training method of the CRNN-LPR character recognition network includes:

[0017] Collect license plate pictures through the field and the network;

[0018] Clean and annotate the license plate pictures to obtain license plate picture data;

[0019] Adopt a generative data augmentation strategy for the license plate picture data to generate a license plate picture training dataset;

[0020] Train the CRNN-LPR character recognition network based on the license plate picture training dataset to obtain a trained CRNN-LPR character recognition network.

[0021] In some embodiments, the generative data augmentation strategy includes:

[0022] Use image rotation and affine transformation techniques to simulate different shooting angles and obtain license plate picture data with different shooting angles;

[0023] Simulate the deformation of the license plate in the actual scene through geometric deformation, where the geometric deformation includes tilting, bending, and stretching, to obtain license plate picture data with different deformations;

[0024] By using the techniques of randomly adjusting brightness and contrast and gamma transformation, the appearance changes of license plates under different lighting conditions are simulated to obtain license plate picture data under different lighting conditions;

[0025] By adding noise, blurring, and raindrop effects, the license plate images under complex weather conditions are simulated to obtain license plate picture data under complex weather conditions;

[0026] The occlusion method is used to simulate the situation where the license plate is occluded to obtain occluded license plate picture data.

[0027] In a second aspect, an embodiment of the present invention further provides a license plate number recognition device, and the device includes:

[0028] An acquisition module, configured to acquire a license plate image to be recognized;

[0029] A positioning module, configured to use the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area, wherein the YOLOv5-MobileNet model is obtained by combining YOLOv5s and MobileNet, the YOLOv5s is used to quickly detect the rough position of the license plate, and the MobileNet accurately locates the license plate area through a lightweight convolutional neural network;

[0030] A correction module, configured to use a spatial transformation network to correct the inclined license plate area to obtain the corrected license plate area;

[0031] An identification module, configured to input the corrected license plate area into a trained CRNN-LPR character recognition network, perform license plate character recognition and extract the license plate number, and output the license plate number recognition result.

[0032] In a third aspect, an embodiment of the present invention further provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the license plate number recognition method described in the first aspect is implemented.

[0033] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, storing computer-executable instructions, and the computer-executable instructions are used to execute the license plate number recognition method described in the first aspect.

[0034] According to the license plate number recognition method and related device provided by the embodiments of the present invention, the license plate number recognition method includes: obtaining a license plate image to be recognized; using the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area, where the YOLOv5-MobileNet model is obtained by combining YOLOv5s and MobileNet, YOLOv5s is used to quickly detect the rough position of the license plate, and MobileNet accurately locates the license plate area through a lightweight convolutional neural network; using a spatial transformation network to correct the tilted license plate area to obtain a corrected license plate area; inputting the corrected license plate area into a trained CRNN-LPR character recognition network to perform license plate character recognition and extract the license plate number, and output the license plate number recognition result. Based on this, the embodiments of the present invention can effectively improve the accuracy and real-time performance of license plate recognition. Description of the Drawings

[0035] Figure 1 is a flowchart of the license plate number recognition method provided by an embodiment of the present invention;

[0036] Figure 2 are the network parameters of the CRNN-LPR network provided by an embodiment of the present invention;

[0037] Figure 3 is a schematic diagram of the network structure of the YOLOv5-MobileNet model provided by an embodiment of the present invention;

[0038] Figure 4 is a license plate data set generated by using a generative data augmentation strategy provided by an embodiment of the present invention;

[0039] Figure 5 is a principle block diagram of the license plate number recognition method provided by an embodiment of the present invention;

[0040] Figure 6 is a schematic diagram of the license plate number recognition device provided by an embodiment of the present invention;

[0041] Figure 7 is a schematic diagram of an electronic device provided by an embodiment of the present invention. Detailed Embodiments

[0042] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0043] It should be noted that although the functional modules are divided in the device schematic diagram and the logical sequence is shown in the flowchart, in some cases, the steps shown or described can be executed in a different module division from that in the device or a different sequence from that in the flowchart. Terms such as "first", "second", etc. in the description, claims and the following drawings are used to distinguish similar objects and do not necessarily have to describe a specific order or sequence.

[0044] In the embodiments of the present invention, words such as "furthermore", "exemplarily" or "optionally" are used to represent examples, illustrations or explanations, and should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Using words such as "furthermore", "exemplarily" or "optionally" aims to present relevant concepts in a specific way.

[0045] For the convenience of subsequently describing the working principle of the embodiments of the present invention, the following first gives an introduction to the related technical scenarios.

[0046] The license plate recognition system is an automated system based on computer vision and artificial intelligence technology, used to recognize the license plate information of vehicles. Its main function is to capture the license plate of a vehicle through an image or video stream and extract the character information therein to achieve the identification and tracking of the vehicle's identity. The license plate recognition system is usually applied to scenarios such as traffic management, parking lot systems, electronic toll collection, and highway monitoring.

[0047] Currently, the license plate detection and recognition technology adopted by the license plate recognition system has technical defects such as inaccurate positioning, low recognition efficiency, and poor real-time performance for the recognition of double-layer license plates. Therefore, how to effectively improve the accuracy and real-time performance of license plate recognition has become an urgent technical problem to be solved.

[0048] Based on this, the present invention provides a license plate number recognition method and related device. Among them, the license plate number recognition method includes: obtaining a license plate image to be recognized; using the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain the license plate area. The YOLOv5-MobileNet model is obtained by combining YOLOv5s and MobileNet. YOLOv5s is used to quickly detect the rough position of the license plate, and MobileNet accurately locates the license plate area through a lightweight convolutional neural network; using a spatial transformation network to correct the tilted license plate area to obtain the corrected license plate area; inputting the corrected license plate area into the trained CRNN-LPR character recognition network to perform license plate character recognition and extract the license plate number, and output the license plate number recognition result. Based on this, the embodiments of the present invention can effectively improve the accuracy and real-time performance of license plate recognition.

[0049] The following further elaborates on the embodiments of the present invention in conjunction with the drawings.

[0050] As Figure 1 shown Figure 1 Figure 1 This is a flowchart of a license plate number recognition method provided by an embodiment of the present invention. By obtaining a license plate image to be recognized; using the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area, where the YOLOv5-MobileNet model is obtained by combining YOLOv5s and MobileNet, YOLOv5s is used to quickly detect the rough position of the license plate, and MobileNet accurately locates the license plate area through a lightweight convolutional neural network; using a spatial transformation network to correct the tilted license plate area to obtain a corrected license plate area; inputting the corrected license plate area into a trained CRNN-LPR character recognition network to perform license plate character recognition and extract the license plate number, and output the license plate number recognition result, thereby realizing license plate number recognition.

[0051] Specifically, the recognition method may include the following steps:

[0052] Step 1: Collect pictures on-site and through the network, and clean and label the collected data.

[0053] Step 2: In view of the problem of insufficient license plate data, adopt a generative data augmentation strategy to generate diverse license plate images and expand the training data set.

[0054] Step 3: In the license plate detection and positioning stage, combine the YOLOv5s and MobileNet algorithms to accurately locate the license plate area in the picture, improving the detection efficiency and accuracy.

[0055] Step 4: Use a spatial transformation network (STN) to correct the tilted license plate image to ensure the standardized processing of the license plate area.

[0056] Step 5: Input the corrected license plate image into the CRNN-LPR character recognition network to perform character recognition and extract the license plate number.

[0057] Step 6: Output the recognition result to obtain the accurate license plate number.

[0058] It can be understood that in step 2, the specific methods of the generative data augmentation strategy include: First, image rotation and affine transformation techniques are used to simulate different shooting angles, and geometric deformations (such as tilting, bending, stretching, etc.) are used to simulate the deformation of license plates in actual scenarios. Furthermore, techniques such as random adjustment of brightness, contrast, and gamma transformation are adopted to simulate the appearance changes of license plates under different lighting conditions. In addition, by adding effects such as noise, blur, and raindrops, license plate images under complex weather conditions are simulated to improve the performance of the model in harsh environments. Finally, occlusion methods are used to simulate the situation where license plates are occluded, improving the recognition ability of the model under partial occlusion. Through these generative data augmentation strategies, the diversity of the dataset is significantly improved, enabling the model to maintain high robustness and accuracy in complex scenarios, especially in areas with scarce data, significantly enhancing the generalization ability of license plate recognition.

[0059] It can be understood that in step 3, for the dataset generated by the generative data augmentation strategy in step 2, by combining the YOLOv5s and MobileNet algorithms, the license plate area in the input image is accurately located. YOLOv5s is used to quickly detect the rough position of the license plate, while MobileNet further improves the positioning accuracy through a lightweight convolutional neural network, thus effectively locating the license plate area and ensuring that the subsequent character recognition stage can be carried out within the accurate license plate area.

[0060] It can be understood that in step 4, the license plate images extracted in step 3 are corrected, especially in the case of license plate images being tilted or deformed, thereby improving the accuracy of license plate detection and recognition.

[0061] It can be understood that in step 5, the corrected license plate images in step 4 are input into the CRNN-LPR character recognition network, and license plate character recognition is performed by combining the improved CRNN structure with LPRNet. This network uses a fully convolutional structure and a CTC loss function, removes the Bi-LSTM module in the traditional CRNN, and further improves the character recognition accuracy through LocNet. The entire network adopts a lightweight design, reduces the size of each layer of feature maps, and outputs the probability distribution of license plate characters through fully convolutional layers. The input of the CRNN-LPR network is the license plate image corrected by STN, and the output is the character sequence of the license plate number. Among them, the network parameters of the CRNN-LPR network are as Figure 2 shown.

[0062] It should be noted that the experimental results in step 5 show that YOLO-MobileNet converges faster, has a smaller weight in the license plate detection task, is more suitable for deployment on embedded devices, and demonstrates stronger generalization ability in the mixed dataset test. CRNN has an inference speed of 21.04 ms while maintaining an accuracy of 86.86%, which can meet the requirements of real-time applications.

[0063] It can be understood that in the license plate detection and positioning stage of the embodiments of the present invention, YOLOv5s is combined with the MobileNet algorithm to locate the license plate part in the picture, and the spatial transformation network STN corrects the tilted license plate, and the corrected license plate is sent to the CRNN-LPR character recognition network for recognition.

[0064] It can be understood that during the training process of the network model in the embodiments of the present invention, in view of the problem of lack of license plate data in some underdeveloped areas, a generative data augmentation strategy is proposed, and at the same time, the license plate detection network and the post-processing method are optimized and improved.

[0065] It can be understood that the network structure of the YOLOv5-MobileNet model in the embodiments of the present invention is as Figure 3 shown. YOLOv5-MobileNetV3 replaces the original backbone network of YOLOv5 to reduce the number of model parameters and accelerate inference. FPN (Feature Pyramid Network) fuses multi-layer features to improve the detection accuracy of license plates. STN (Spatial Transformation Network) performs geometric transformation correction on tilted license plates to reduce recognition errors. CRNN-LPRNet combines the end-to-end recognition advantage of CRNN and the lightweight structure of LPRNet, retains the end-to-end image processing of CRNN, uses a fully convolutional network instead of Bi-LSTM, and adds a CTC loss function to improve the recognition accuracy and real-time performance.

[0066] It can be understood that the embodiments of the present invention introduce generative data augmentation strategies such as perspective transformation, distortion, and noise addition, and the generated license plate dataset is as Figure 4As shown below. The specific methods of the generative data augmentation strategy include: First, use image rotation and affine transformation techniques to simulate different shooting angles, and simulate the deformation of license plates in actual scenarios through geometric deformations (such as tilting, bending, stretching, etc.); Secondly, adopt techniques such as random adjustment of brightness, contrast, and gamma transformation to simulate the appearance changes of license plates under different lighting conditions. In addition, simulate license plate images under complex weather conditions by adding effects such as noise, blur, and raindrops to improve the performance of the model in harsh environments; Finally, use occlusion methods to simulate the situation where license plates are occluded, and improve the recognition ability of the model under partial occlusion. Through these generative data augmentation strategies, the diversity of the dataset is significantly improved, enabling the model to maintain high robustness and accuracy in complex scenarios. Especially in areas with scarce data, the generalization ability of license plate recognition is significantly enhanced.

[0067] It can be understood that as Figure 5 shown in the principle block diagram, the license plate number recognition method of the present invention uses MobileNetV3 to replace the original backbone network of YOLOv5 to locate the license plate part in the picture. FPN (Feature Pyramid Network) fuses multi-layer features to improve the detection accuracy of license plates and outputs multi-scale fusion features. STN (Spatial Transformer Network) corrects the geometric transformation of tilted license plates and sends them to the CRNN character recognition network for recognition. CRNN-LPRNet combines the end-to-end recognition advantages of CRNN and the lightweight structure of LPRNet, retains the end-to-end image processing of CRNN, uses a fully convolutional network instead of Bi-LSTM, and adds a CTC loss function to output the license plate number.

[0068] It should be noted that the present invention constructs a license plate dataset for verification experiments, and the recognition rate of license plate numbers reaches 86.86%. Moreover, the size of the network model proposed by the present invention is less than 21.64MB, with strong real-time performance. Experiments prove that the YOLO-CRNN model can achieve a recognition accuracy of 93.38% and a speed of 30fps in practical applications, and has better detection accuracy compared with YOLO, CRNN, and LPR. The speed can meet the requirements of real-time license plate detection and recognition of moving vehicles in intelligent transportation systems.

[0069] Based on this, a license plate number recognition method proposed by the present invention includes: obtaining a license plate image to be recognized; using the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area, where the YOLOv5-MobileNet model is obtained by combining YOLOv5s and MobileNet, YOLOv5s is used to quickly detect the rough position of the license plate, and MobileNet accurately locates the license plate area through a lightweight convolutional neural network; using a spatial transformation network to correct the tilted license plate area to obtain a corrected license plate area; inputting the corrected license plate area into a trained CRNN-LPR character recognition network to perform license plate character recognition and extract the license plate number, and output the license plate number recognition result. Based on this, the embodiments of the present invention can effectively improve the accuracy and real-time performance of license plate recognition.

[0070] In addition, as Figure 6 shown, an embodiment of the present invention also discloses a license plate number recognition device, which includes:

[0071] An acquisition module 110, configured to acquire a license plate image to be recognized;

[0072] A positioning module 120, configured to use the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area, where the YOLOv5-MobileNet model is obtained by combining YOLOv5s and MobileNet, YOLOv5s is used to quickly detect the rough position of the license plate, and MobileNet accurately locates the license plate area through a lightweight convolutional neural network;

[0073] A correction module 130, configured to use a spatial transformation network to correct the tilted license plate area to obtain a corrected license plate area;

[0074] An identification module 140, configured to input the corrected license plate area into a trained CRNN-LPR character recognition network to perform license plate character recognition and extract the license plate number, and output the license plate number recognition result.

[0075] The license plate number recognition device in the embodiments of the present invention is used to execute the license plate number recognition method in the above embodiments, and its specific processing process is the same as that of the license plate number recognition method in the above embodiments, and will not be elaborated here one by one.

[0076] In addition, as Figure 7As shown in the figure, an embodiment of the present invention also discloses an electronic device, including: at least one processor 210; at least one memory 220 for storing at least one program; when the at least one program is executed by the at least one processor 210, the license plate number recognition method in any of the previous embodiments is implemented.

[0077] In addition, an embodiment of the present invention also discloses a computer-readable storage medium, in which computer-executable instructions are stored, and the computer-executable instructions are used to execute the license plate number recognition method in any of the previous embodiments.

[0078] The system architecture and application scenarios described in the embodiments of the present invention are for more clearly illustrating the technical solutions of the embodiments of the present invention, and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. Those skilled in the art know that with the evolution of the system architecture and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present invention are equally applicable to similar technical problems.

[0079] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations.

[0080] In the hardware implementation, the division between the functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component can have multiple functions, or a function or step can be executed by several physical components in cooperation. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or can be implemented as hardware, or can be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, and the computer-readable medium can include a computer storage medium (or a non-transitory medium) and a communication medium (or a transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory, or other memory technologies, CD-ROM, digital versatile disc (DVD), or other optical disc storage, magnetic cassette, tape, magnetic disk storage, or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, a communication medium typically includes computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0081] As used in this specification, the terms "component", "module", "system", etc. are used to refer to computer-related entities, hardware, firmware, combinations of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable, an execution thread, a program, or a computer. By way of illustration, both an application running on a computing device and the computing device can be components. One or more components can reside within a process or execution thread, and a component can be located on one computer or distributed between two or more computers. Further, these components can execute from various computer-readable media having various data structures stored thereon. A component can, for example, communicate by way of signals with other systems, such as via a local or remote process, in accordance with one or more data packets (e.g., data from two components interacting with one another from a local system, a distributed system, or a network, such as the Internet).

Claims

1. A license plate number recognition method, characterized in that, Including: Obtain a license plate image to be recognized; Use the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area. Among them, the YOLOv5-MobileNet model is obtained by combining YOLOv5s and MobileNet. YOLOv5s is used to quickly detect the rough position of the license plate, and MobileNet accurately locates the license plate area through a lightweight convolutional neural network; Use a spatial transformation network to correct the tilted license plate area to obtain the corrected license plate area; Input the corrected license plate area into the trained CRNN-LPR character recognition network to perform license plate character recognition and extract the license plate number, and output the license plate number recognition result.

2. The license plate number recognition method according to claim 1, characterized in that, The step of using the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area includes: Replace the original backbone network in YOLOv5 with MobileNetV3 to locate the license plate part in the license plate image to obtain a license plate area.

3. The license plate number recognition method according to claim 1, wherein After using the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area, it further includes: Use a feature pyramid network to fuse multi-layer features of the license plate area and output multi-scale fusion features.

4. The license plate number recognition method according to claim 1, characterized in that, The CRNN-LPR character recognition network includes an improved CRNN, LPRNet, and LocNet. The CRNN-LPR character recognition network uses a fully convolutional layer network and adds a CTC loss function. The fully convolutional network replaces the Bi-LSTM module in the CRNN, and the probability distribution of license plate characters is output through the fully convolutional layer. LocNet is used to improve the accuracy of license plate character recognition.

5. The license plate number recognition method according to claim 1, characterized in that The YOLOv5-MobileNet model is deployed on an embedded device.

6. The license plate number recognition method according to claim 1, characterized in that, The training method of the CRNN-LPR character recognition network includes: Collect license plate pictures through the field and the network; Clean and annotate the license plate pictures to obtain license plate picture data; Adopt a generative data augmentation strategy for the license plate picture data to generate a license plate picture training dataset; Train the CRNN-LPR character recognition network based on the license plate picture training dataset to obtain a trained CRNN-LPR character recognition network.

7. The license plate number recognition method according to claim 6, characterized in that, The generative data augmentation strategy includes: Use image rotation and affine transformation techniques to simulate different shooting angles to obtain license plate picture data with different shooting angles; Simulate the deformation of the license plate in the actual scene through geometric deformation, where the geometric deformation includes tilting, bending, and stretching, to obtain license plate picture data with different deformations; Adopt techniques of randomly adjusting brightness, contrast, and gamma transformation to simulate the appearance changes of the license plate under different lighting conditions to obtain license plate picture data under different lighting conditions; Simulate the license plate image under complex weather conditions by adding noise, blur, and raindrop effects to obtain license plate picture data under complex weather conditions; Use the occlusion method to simulate the situation where the license plate is occluded, and obtain the occluded license plate image data.

8. A license plate number recognition device, characterized in that, The device includes: An acquisition module, configured to acquire a license plate image to be recognized; A positioning module, configured to use the YOLOv5-MobileNet model to locate the license plate part in the license plate image to obtain a license plate area, wherein the YOLOv5-MobileNet model is obtained by combining YOLOv5s and MobileNet, the YOLOv5s is used to quickly detect the rough position of the license plate, and the MobileNet accurately locates the license plate area through a lightweight convolutional neural network; A correction module, configured to use a spatial transformation network to correct the tilted license plate area to obtain the corrected license plate area; An identification module, configured to input the corrected license plate area into a trained CRNN-LPR character recognition network, perform license plate character recognition and extract the license plate number, and output the license plate number recognition result.

9. An electronic device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the license plate number recognition method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing computer-executable instructions for executing the license plate number recognition method according to any one of claims 1 to 7.

Images