License plate attribute recognition method, device and storage medium

Abstract

The present application discloses a method, device and storage medium for recognizing license plate attributes, wherein the method for recognizing license plate attributes includes: performing multiple rounds of license plate attribute estimation processing on a license plate image to be recognized, and obtaining multiple license plate attribute estimation results of a target license plate; in one round of license plate attribute estimation processing: performing feature extraction on the license plate image to be recognized, obtaining a target attribute feature of the target license plate, and determining one of the multiple license plate attribute estimation results based on the target attribute feature and the correlation between each candidate license plate attribute category in a candidate license plate attribute category set; the candidate license plate attribute category set corresponds to the round of license plate attribute estimation processing; and determining the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results. The present application obtains multiple license plate attribute estimation results through multiple rounds of license plate attribute estimation processing, and determines the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results, thereby improving the accuracy of license plate attribute recognition.

Description

Technical Field

[0001] The present application belongs to the field of image processing technology, and specifically relates to a method, device and storage medium for recognizing license plate attributes. Background Art

[0002] License plate attribute recognition is mainly about classifying license plates. At the global level, license plates need to be distinguished between different countries. Within a country, each region needs to be distinguished to facilitate vehicle management and smooth traffic. The technology involved is mainly a deep learning recognition and classification network.

[0003] At present, how to improve the accuracy of license plate attribute recognition is an urgent problem to be solved. Summary of the invention

[0004] The present application provides a method, device and storage medium for recognizing license plate attributes to improve the accuracy of license plate attribute recognition.

[0005] To solve the above technical problems, a technical solution adopted in the present application is: a method for identifying license plate attributes, the identification method comprising: performing multiple rounds of license plate attribute estimation processing on the license plate image to be identified, and obtaining multiple license plate attribute estimation results of the target license plate; the target license plate includes the license plate in the license plate image to be identified; in one round of the license plate attribute estimation processing: performing feature extraction on the license plate image to be identified, obtaining the target attribute feature of the target license plate, and determining one of the multiple license plate attribute estimation results based on the correlation between the target attribute feature and each candidate license plate attribute category in the candidate license plate attribute category set; the candidate license plate attribute category set corresponds to this round of license plate attribute estimation processing; based on the multiple license plate attribute estimation results, determining the license plate attribute category of the target license plate.

[0006] To solve the above technical problems, another technical solution adopted in the present application is: an electronic device, comprising a memory and a processor coupled to each other, wherein the processor is used to execute program instructions stored in the memory to implement any of the above identification methods.

[0007] In order to solve the above technical problems, another technical solution adopted by the present application is: a computer-readable storage medium, on which program data is stored, and when the program data is executed by a processor, any of the above identification methods is implemented.

[0008] The present application has at least the following advantages: the recognition method of the present application obtains multiple license plate attribute estimation results through multiple rounds of license plate attribute estimation processing, and determines the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results, which can improve the accuracy of license plate attribute recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0010] Figure 1 It is a flow chart of an embodiment of an attribute recognition sub-model training method in the license plate attribute recognition method of the present application;

[0011] Figure 2 It is a flowchart of an embodiment of a method for recognizing license plate attributes of the present application;

[0012] Figure 3 It is a schematic diagram of the framework of an embodiment of a device for recognizing license plate attributes of the present application;

[0013] Figure 4 It is a schematic diagram of a framework of an electronic device according to an embodiment of the present application;

[0014] Figure 5 It is a schematic diagram of a framework of an embodiment of a computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0016] See also Figure 1 , Figure 1 It is a flow chart of an embodiment of an attribute recognition sub-model training method in the license plate attribute recognition method of the present application.

[0017] An embodiment of the present application provides an attribute recognition sub-model training method, comprising the following steps:

[0018] S11: Obtain a training license plate image set of multiple candidate license plate attribute categories.

[0019] In the training phase, the training license plate images are first classified into attribute categories based on the candidate license plate attribute categories, which include the color of the license plate, the region to which the license plate belongs, etc. The region to which the license plate belongs includes the country to which the license plate belongs or the region to which the license plate belongs, etc. The training license plate images belonging to the same candidate license plate attribute category are classified into the same training license plate image.

[0020] It should be noted that obtaining a training license plate image set of multiple candidate license plate attribute categories also includes:

[0021] A training license plate image is obtained, and the training license plate image is captured by a camera. The training license plate image is annotated with candidate license plate attribute categories, license plate frames, and license plate character information to obtain an annotated training license plate image.

[0022] S12: Clustering the training license plate image sets of multiple candidate license plate attribute categories to obtain at least two fused training image sets.

[0023] Based on research, the applicant found that due to the imbalance of development between countries and regions, people's income and economic level determine the number of vehicles owned, so the distribution of license plate attribute categories is very uneven. In some countries and regions, there are a lot of license plates, while in other regions, there are relatively few. The uneven distribution of the number of training data in different training license plate image sets will have a great impact on the accuracy of the recognition of license plate attribute categories directly used for training attribute recognition sub-models.

[0024] In this embodiment, each fused training image set includes at least two training license plate images of candidate license plate attribute categories, and the difference in the number of training license plate images of different candidate license plate attribute categories in each fused training image set is less than the difference threshold. The difference threshold is determined according to the specific number of training license plate images, based on the balance of data volume and without affecting the accuracy of the candidate license plate attribute category.

[0025] Specifically, clustering is performed on a plurality of training license plate image sets of candidate license plate attribute categories to obtain at least two fused training image sets, including:

[0026] Sort all training license plate image sets in order of data volume.

[0027] Determine the multiples of the training license plate image set of each type of data volume and the training license plate image set of the minimum data volume.

[0028] The training license plate image sets with the same integer part of the multiples are fused into a fused training image set.

[0029] The fused training image set containing only one training license plate image set is merged into the adjacent multiple fused training image sets.

[0030] A specific implementation is described:

[0031] Arrange the original M training license plate image sets in ascending order according to the amount of data {C1, C2, C3, ... C m}. Calculate the maximum amount of labeled data set C m The integer part N of the multiple of the minimum amount of labeled data set C1m . Calculate the second largest label dataset C m-1 The integer part N of the multiple of the minimum amount of labeled data set C1 m-1 By analogy, the training license plate image sets of various data volumes and the minimum data volume C are calculated. m A multiple of the number of training license plate image sets.

[0032] The training license plate image sets with the same value of N are merged into a fused training image set, so that the difference in the number of training license plate images of different candidate license plate attribute categories in each fused training image set is less than the difference threshold.

[0033] However, since the data volume of the training license plate images in each training license plate image set is different, after classification according to the above method, there is a fused training image set that only includes one training license plate image set, which cannot be used alone to train the attribute recognition sub-model, and needs to be merged into a fused training image set of adjacent multiples N including at least two training license plate image sets. By merging into fused training image sets of adjacent multiples, the difference in the number of training license plate images of different candidate license plate attribute categories in each fused training image set is less than the difference threshold, which does not affect the accuracy of the attribute recognition sub-model.

[0034] In other embodiments, in addition to the above-mentioned multiplicative clustering method, other methods may be used to form a fused training image set so that the difference in the number of training license plate images of different candidate license plate attribute categories in each fused training image set is less than a difference threshold.

[0035] S13: Using at least two fused training image sets to train a corresponding number of attribute recognition sub-models to obtain at least two attribute recognition sub-models.

[0036] A corresponding number of attribute recognition sub-models are trained using at least two fused training image sets to obtain at least two attribute recognition sub-models. It should be noted that different attribute recognition sub-models in the obtained at least two attribute recognition sub-models have the same structure and different model parameters, that is, the network structure of the same neural network can be used to construct at least two attribute recognition sub-models before training, and the at least two attribute recognition sub-models are trained using different fused training image sets to obtain at least two attribute recognition sub-models with the same structure but different model parameters.

[0037] This method does not require complex data processing methods such as data enhancement and data balancing, and can achieve that the difference in the number of training license plate images of different candidate license plate attribute categories in each fusion training image set is less than the difference threshold. The attribute recognition sub-model trained by this training method has a high recognition accuracy rate for license plate attribute categories.

[0038] See also Figure 2, Figure 2 It is a flowchart of an embodiment of a method for recognizing license plate attributes of the present application.

[0039] Another embodiment of the present application provides a method for identifying license plate attributes, comprising the following steps:

[0040] S21: Perform multiple rounds of license plate attribute estimation processing on the license plate image to be recognized to obtain multiple license plate attribute estimation results of the target license plate.

[0041] By performing multiple rounds of license plate attribute estimation processing on the license plate image to be recognized, multiple license plate attribute estimation results of the target license plate can be obtained. The multiple rounds of license plate attribute estimation processing can be performed synchronously or sequentially.

[0042] The target license plate includes the license plate in the license plate image to be recognized.

[0043] In a round of license plate attribute estimation processing, it includes: extracting features of the license plate image to be recognized, obtaining target attribute features of the target license plate, and determining one of the multiple license plate attribute estimation results based on the target attribute features and the degree of association between each candidate license plate attribute category in the candidate license plate attribute category set. The candidate license plate attribute category set corresponds to this round of license plate attribute estimation processing.

[0044] It should be noted that the candidate license plate attribute category set in each round of license plate attribute estimation processing includes at least two candidate license plate attribute categories. Usually, the candidate license plate attributes corresponding to different rounds of license plate attribute estimation processing can be completely different or partially the same, and even the candidate license plate attribute categories of some rounds of license plate attribute estimation processing can be the same.

[0045] Specifically, performing multiple rounds of license plate attribute estimation processing on the license plate image to be recognized to obtain multiple license plate attribute estimation results of the target license plate includes:

[0046] The license plate image to be identified is input into multiple attribute recognition sub-models to obtain multiple license plate attribute estimation results. Among them, different attribute recognition sub-models have the same structure and different model parameters, that is, the same neural network is used and trained through different fusion training image sets. One attribute recognition sub-model is used to perform a round of license plate attribute estimation processing on the license plate image to be identified, and obtain one of the multiple license plate attribute estimation results.

[0047] Specifically, the attribute recognition sub-model can be obtained by training using the training method in the previous embodiment.

[0048] Among them, in a round of license plate attribute estimation processing, the correlation between the target attribute feature and each candidate license plate attribute category in the candidate license plate attribute category set is the confidence that the license plate image to be identified is identified as each candidate license plate attribute category. The license plate attribute estimation result includes the confidence of a candidate license plate attribute category in the candidate license plate attribute category set. Usually, the candidate license plate attribute category with the highest confidence is selected as the license plate attribute estimation result of this round of license plate attribute estimation processing.

[0049] S22: Determine the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results.

[0050] In one embodiment, determining the license plate attribute category of the target license plate based on multiple license plate attribute estimation results includes: determining the license plate attribute estimation result with the highest confidence among multiple license plate attribute estimation results obtained by multiple rounds of license plate attribute estimation processing; and determining the candidate license plate attribute category corresponding to the determined license plate attribute estimation result as the license plate attribute category of the target license plate.

[0051] Therefore, the recognition method of the present application obtains multiple license plate attribute estimation results through multiple rounds of license plate attribute estimation processing, and determines the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results, which can improve the accuracy of license plate attribute recognition.

[0052] The license plate attribute categories include the license plate color or the region to which the license plate belongs, and the region to which the license plate belongs includes the country or region to which it belongs, etc.

[0053] Specifically, determining the license plate attribute estimation result with the highest confidence among the multiple license plate attribute estimation results includes: based on the multiple license plate attribute estimation results, determining the license plate attribute estimation result with the highest confidence from the license plate attribute estimation results with confidence greater than a threshold, thereby ensuring the reliability of the license plate attribute estimation result with the highest confidence finally selected, and improving the accuracy of license plate attribute recognition.

[0054] In one embodiment, after determining the license plate attribute category of the target license plate based on the plurality of license plate attribute estimation results, the method further includes:

[0055] S23: Based on the license plate image to be recognized, character information of the target license plate is obtained.

[0056] Based on the license plate image to be recognized, character information of the target license plate is obtained, and the character information includes special characters and / or character length. Among them, special characters include provincial capital abbreviations and the like.

[0057] It should be noted that the character information of the target license plate can be obtained by inputting the image to be recognized into the character recognition model to obtain the character information recognition result. The character recognition model is obtained by training with a training license plate character image set. Of course, if the amount of data between different training license plate character image sets is unbalanced, the method in the above embodiment can also be used to form a fused training character image set, so that the difference in the amount of data belonging to different training license plate character image sets in the fused training character image set is less than the difference threshold. Then use the fused training character image set to train multiple character recognition models for multiple rounds of character information estimation processing of the license plate image to be recognized. The specific process will not be repeated here.

[0058] S24: If it is determined that the character information and the license plate attribute category of the target license plate do not match, then the license plate attribute category of the target license plate is re-determined based on the candidate license plate attribute category corresponding to the character information.

[0059] If it is determined that the character information and the license plate attribute category of the target license plate do not match, the license plate attribute category of the target license plate is re-determined based on the candidate license plate attribute category corresponding to the character information.

[0060] In one embodiment, determining whether the character information matches the license plate attribute category of the target license plate includes:

[0061] Compare the character information with the character features corresponding to the license plate attribute category.

[0062] If the comparison is consistent, it is determined that the character information matches the license plate attribute category, and there is no need to redetermine the license plate attribute category of the target license plate.

[0063] If the comparison does not match, it is determined that the character information does not match the license plate attribute category, and the license plate attribute category of the target license plate needs to be re-determined.

[0064] Herein, a specific implementation method is used to illustrate how to compare character information with character features corresponding to the license plate attribute category:

[0065] According to the license plate attribute category obtained in step S22, the corresponding character features in the preset information table are queried (the preset information table is pre-prepared according to the special characters of the license plate and the prior information of the character length); it is determined whether the comparison result is consistent with the character information obtained in step S23. For example, if the license plate attribute category of step S22 is country A, then a single comparison is made to see whether the character features in the preset information table of country A are consistent with the character information in step S23. If they are consistent, it means that the character information matches the license plate attribute category; if they are not consistent, it means that there is an error in the license plate attribute category (the error may be caused by the error in the annotation of the training license plate image during training), and the license plate attribute category is corrected using the character information. Of course, in other embodiments, other comparison methods can also be used to verify whether there are obvious errors in the license plate attribute category.

[0066] Using character information to correct the license plate attribute category can avoid obvious errors in the license plate attribute category. It also reflects the idea of ​​integrating multi-dimensional information of license plate character information in the process of license plate attribute recognition to improve the accuracy.

[0067] In summary, the license plate attribute recognition method of the present application obtains multiple license plate attribute estimation results through multiple rounds of license plate attribute estimation processing, and determines the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results, which can improve the accuracy of license plate attribute recognition. In addition, in this application, the data distribution is balanced to form multiple attribute recognition sub-models for multiple rounds of license plate attribute estimation processing to improve the accuracy of license plate attribute recognition. In addition, the license plate attribute category is corrected by using character information in this application, which can avoid obvious errors in the license plate attribute category, and also reflects the idea of ​​integrating the multi-dimensional information joint recognition of license plate character information in the license plate attribute recognition process to improve the accuracy.

[0068] See also Figure 3 , Figure 3 It is a schematic diagram of the framework of an embodiment of a device for recognizing license plate attributes of the present application.

[0069] Another embodiment of the present application provides a license plate attribute recognition device 30, including multiple license plate attribute estimation processing modules 31 and an attribute category determination module 32. The multiple license plate attribute estimation processing modules 31 perform multiple rounds of license plate attribute estimation processing on the license plate image to be identified, and obtain multiple license plate attribute estimation results of the target license plate. The target license plate includes the license plate in the license plate image to be identified.

[0070] In a round of license plate attribute estimation processing, the license plate attribute estimation processing module 31 extracts features of the license plate image to be identified, obtains target attribute features of the target license plate, and determines one of the multiple license plate attribute estimation results based on the target attribute features and the degree of association between each candidate license plate attribute category in the candidate license plate attribute category set. The candidate license plate attribute category set corresponds to this round of license plate attribute estimation processing.

[0071] The attribute category determination module 32 determines the license plate attribute category of the target license plate based on a plurality of license plate attribute estimation results.

[0072] The recognition device 30 of the present application can implement the method for recognizing license plate attributes in any of the above-mentioned embodiments.

[0073] The license plate attribute recognition device 30 of the present application obtains multiple license plate attribute estimation results through multiple rounds of license plate attribute estimation processing, and determines the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results, thereby improving the accuracy of license plate attribute recognition.

[0074] See also Figure 4 , Figure 4 It is a schematic diagram of the framework of an embodiment of an electronic device of the present application.

[0075] Another embodiment of the present application provides an electronic device 40, including a memory 41 and a processor 42 coupled to each other, the processor 42 is used to execute program instructions stored in the memory 41 to implement the license plate attribute recognition method of any of the above embodiments. In a specific implementation scenario, the electronic device 40 may include but is not limited to: a microcomputer, a server, and in addition, the electronic device 40 may also include a mobile device such as a laptop computer and a tablet computer, which is not limited here.

[0076] Specifically, the processor 42 is used to control itself and the memory 41 to implement the steps in any of the above-mentioned embodiments. The processor 42 can also be called a CPU (Central Processing Unit). The processor 42 may be an integrated circuit chip with signal processing capabilities. The processor 42 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field-programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. In addition, the processor 42 can be implemented by an integrated circuit chip.

[0077] See also Figure 5 , Figure 5 It is a schematic diagram of a framework of an embodiment of a computer-readable storage medium of the present application.

[0078] Another embodiment of the present application provides a computer-readable storage medium 50 on which program data 51 is stored. When the program data 51 is executed by a processor, the method for recognizing license plate attributes of any of the above embodiments is implemented.

[0079] In the several embodiments provided in the present application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device implementation described above is only schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation, such as units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0080] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0081] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0082] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium 50. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a readable storage medium 50, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of each implementation method of the present application. The aforementioned readable storage medium 50 includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0083] The above descriptions are merely embodiments of the present application and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for recognizing license plate attributes, characterized in that: The identification method comprises: The license plate image to be identified is subjected to multiple rounds of license plate attribute estimation processing to obtain multiple license plate attribute estimation results of the target license plate; the target license plate includes the license plate in the license plate image to be identified; in one round of the license plate attribute estimation processing: Extracting features of the license plate image to be identified to obtain target attribute features of the target license plate, and determining a license plate attribute estimation result from the plurality of license plate attribute estimation results based on the degree of association between the target attribute features and each candidate license plate attribute category in a candidate license plate attribute category set; the candidate license plate attribute category set corresponds to the license plate attribute estimation process of this round; Determining the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results; The method performs multiple rounds of license plate attribute estimation processing on the license plate image to be identified to obtain multiple license plate attribute estimation results of the target license plate, including: Input the license plate image to be identified into multiple attribute recognition sub-models to obtain the multiple license plate attribute estimation results; different attribute recognition sub-models have the same structure and different model parameters; one attribute recognition sub-model is used to perform a round of license plate attribute estimation processing on the license plate image to be identified to obtain one license plate attribute estimation result among the multiple license plate attribute estimation results; The multiple attribute recognition sub-models are trained by the following steps: Obtain a training license plate image set of multiple candidate license plate attribute categories; Clustering the training license plate image sets of the multiple candidate license plate attribute categories to obtain at least two fused training image sets; each of the fused training image sets includes training license plate images of at least two candidate license plate attribute categories, and the difference in the number of training license plate images of different candidate license plate attribute categories in each fused training image set is less than a difference threshold; The at least two fused training image sets are used to train a corresponding number of attribute recognition sub-models to obtain at least two attribute recognition sub-models.

2. The method according to claim 1, characterized in that The license plate attribute estimation result includes the confidence of a candidate license plate attribute category in the candidate license plate attribute category set, and determining the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results includes: Determine the license plate attribute estimation result with the highest confidence among the multiple license plate attribute estimation results; and determine the candidate license plate attribute category corresponding to the determined license plate attribute estimation result as the license plate attribute category of the target license plate.

3. The identification method according to any one of claims 1-2, characterized in that: After determining the license plate attribute category of the target license plate based on the multiple license plate attribute estimation results, the method further includes: Based on the license plate image to be recognized, obtaining character information of the target license plate; If it is determined that the character information and the license plate attribute category of the target license plate do not match, the license plate attribute category of the target license plate is re-determined based on the candidate license plate attribute category corresponding to the character information.

4. The identification method according to claim 3, characterized in that: Before determining that the character information does not match the license plate attribute category of the target license plate, the method further includes: Comparing the character information with the character features corresponding to the license plate attribute category; If the comparison is consistent, it is determined that the character information matches the license plate attribute category; If the comparison does not match, it is determined that the character information does not match the license plate attribute category.

5. The identification method according to claim 2, characterized in that: The determining of the license plate attribute estimation result with the highest confidence among the multiple license plate attribute estimation results includes: Based on the multiple license plate attribute estimation results, a license plate attribute estimation result with the highest confidence is determined from the license plate attribute estimation results with confidence greater than a threshold.

6. The identification method according to any one of claims 1 to 2, characterized in that: The candidate license plate attribute categories include license plate color or the region to which the license plate belongs.

7. The identification method according to claim 1, characterized in that: The clustering of the training license plate image sets of the plurality of candidate license plate attribute categories to obtain at least two fused training image sets comprises: Sorting all the training license plate image sets in order of data volume; Determine the multiples of the training license plate image set of each type of data volume and the training license plate image set of the minimum data volume; Merging the training license plate image sets having the same integer part of the multiple into a fused training image set; The fused training image set containing only one of the training license plate image sets is merged into the fused training image set of adjacent multiples.

8. The identification method according to claim 7, characterized in that: The step of obtaining a training license plate image set for each candidate license plate attribute category comprises: The training license plate images are classified into attribute categories based on the candidate license plate attribute categories, and the training license plate images belonging to the same candidate license plate attribute category are classified into the same training license plate image set.

9. An electronic device, characterized in that: The invention comprises a memory and a processor coupled to each other, wherein the processor is used to execute program instructions stored in the memory to implement the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having program data stored thereon, characterized in that: When the program data is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images