WO2020098250A1

WO2020098250A1 - Character recognition method, server, and computer readable storage medium

Info

Publication number: WO2020098250A1
Application number: PCT/CN2019/088638
Authority: WO
Inventors: 许洋; 王健宗; 肖京
Original assignee: 平安科技（深圳）有限公司
Priority date: 2018-11-12
Filing date: 2019-05-27
Publication date: 2020-05-22
Also published as: CN109685100A

Abstract

The present application relates to artificial intelligence. Disclosed is a character recognition method, comprising: obtaining character data and performing image synthesis on each of the obtained character data and a preset background picture to obtain character images corresponding to each of the character data; performing random disturbance processing on the synthesized character images to obtain different types of character images; inputting the different types of character images to a deep learning network for training to generate a character recognition model; and inputting a character image to be recognized to the character recognition model to output a recognition result of said character image. The present application also provides a server and a computer readable storage medium. The character recognition method, server, and computer readable storage medium provided by the present application implement an OCR function on the basis of a deep learning algorithm, and can increase the range of character recognition and improve the accuracy of character recognition.

Description

Character recognition method, server and computer readable storage medium

Priority declaration

This application is based on the declaration of the Paris Convention and enjoys the priority of the Chinese patent application with the application number CN201811341729.X and the name "character recognition method, server and computer-readable storage medium" submitted on November 12, 2018 The entire content is incorporated in this application by reference.

Technical field

The present application relates to the field of character recognition, and in particular, to a character recognition method, server, and computer-readable storage medium.

Background technique

In the process of OCR (Optical Character Recognition, optical character recognition) business, usually in accordance with the needs of the business party to identify certain fields in a specific scene, this generally requires the business party to provide real picture data in the scene, And you need to manually label the data, and then use these labeled pictures to detect and recognize the deep learning training of the model. When the identification content of these fields is in a small limited set (such as the gender of the ID card, the vehicle type of the driving license, the nature of use, etc.), the recognition accuracy rate is usually relatively high. When the identification content of the field is in a very limited set, and even can be regarded as an infinite set (such as the name of the ID card, the owner of the driving license, etc.), the identification is easily limited by the amount of labeled data, accurate The rate will also be affected to a certain extent.

Summary of the invention

In view of this, this application proposes a character recognition method, which can increase the character recognition range and increase the accuracy of character recognition.

First, in order to achieve the above object, a first aspect of the present application provides a server, the server includes a memory and a processor, and the memory stores a character recognition system operable on the processor When executed by the processor, the following steps are implemented:

Obtain character data, and synthesize each acquired character data with a preset background picture to obtain a character image corresponding to each character data;

Randomly disturb the synthesized character image to obtain different types of character images;

Input the different types of character images into the deep learning network for training to generate a character recognition model; and

The character image to be recognized is input into the character recognition model, and the recognition result of the character image to be recognized is output.

In addition, in order to achieve the above object, a second aspect of the present application also provides a character recognition method, which is applied to a server, and the method includes:

Further, in order to achieve the above object, the third aspect of the present application further provides a computer-readable storage medium storing a character recognition system, the character recognition system may be executed by at least one processor, Causing the at least one processor to perform the steps of the character recognition method described in any one of the above.

Compared with the prior art, the character recognition method, server and computer-readable storage medium proposed in this application acquire character data, and perform image synthesis on each character data acquired with a preset background picture to obtain each character data Corresponding character images; performing random perturbation processing on the synthesized character images to obtain different types of character images; inputting the different types of character images into a deep learning network for training to generate a character recognition model; The image is input into the character recognition model, and the recognition result of the character image to be recognized is output. In this way, a variety of training sample data can be generated as needed to solve the problem of small character recognition range and low accuracy due to uneven distribution of real data of training samples in the prior art, increasing the character recognition range and increasing character recognition accuracy .

BRIEF DESCRIPTION

1 is a schematic diagram of an optional hardware architecture of the server of this application;

2 is a schematic diagram of a program module of the first embodiment of the character recognition system of the present application;

3 is a schematic diagram of a program module of the second embodiment of the character recognition system of the present application;

4 is a schematic diagram of an implementation process of the first embodiment of the character recognition method of the present application;

5 is a schematic diagram of an implementation process of a second embodiment of a character recognition method of the present application.

Reference mark:

服务器server	22
网络The internet	33

存储器Memory	1111
处理器processor	1212
网络接口Network Interface	1313
字符识别系统 Character recognition system	100100
获取模块Get module	101101
处理模块Processing module	102102
生成模块Generate module	103103
输出模块Output module	104104
测试模块Test module	105105
调整模块Adjustment module	106106

The implementation, functional characteristics and advantages of the present application will be further described in conjunction with the embodiments and with reference to the drawings.

detailed description

In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be described in further detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

It should be noted that the descriptions related to "first", "second", etc. in this application are for descriptive purposes only, and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated . Thus, the features defined with "first" and "second" may include at least one of the features either explicitly or implicitly. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of those skilled in the art to realize. When the combination of technical solutions contradicts or cannot be realized, it should be considered that the combination of such technical solutions does not exist , Nor within the scope of protection required by this application.

Referring to FIG. 1, it is a schematic diagram of an optional hardware architecture of the application server 2 of the present application.

In this embodiment, the application server 2 may include, but is not limited to, the memory 11, the processor 12, and the network interface 13 may be connected to each other through a system bus. It should be noted that FIG. 2 only shows the application server 2 having the components 11-13, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

The application server 2 may be a computing device such as a rack server, a blade server, a tower server, or a rack server. The application server 2 may be an independent server or a server cluster composed of multiple servers .

The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a hard disk, a multimedia card, a card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static Random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the storage 11 may be an internal storage unit of the application server 2, such as a hard disk or a memory of the application server 2. In other embodiments, the memory 11 may also be an external storage device of the application server 2, such as a plug-in hard disk equipped on the application server 2, a smart memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card (Flash Card), etc. Of course, the memory 11 may also include both the internal storage unit of the application server 2 and its external storage device. In this embodiment, the memory 11 is generally used to store an operating system installed in the application server 2 and various application software, such as program codes of the character recognition system 100. In addition, the memory 11 can also be used to temporarily store various types of data that have been output or will be output.

The processor 12 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip in some embodiments. The processor 12 is generally used to control the overall operation of the application server 2. In this embodiment, the processor 12 is used to run the program code or process data stored in the memory 11, for example, to run the character recognition system 100.

The network interface 13 may include a wireless network interface or a wired network interface. The network interface 13 is generally used to establish a communication connection between the application server 2 and other electronic devices.

So far, the hardware structure and functions of the relevant equipment of this application have been introduced in detail. In the following, various embodiments of the present application will be filed based on the above introduction.

First, the present application proposes a character recognition system 100.

Referring to FIG. 2, it is a program module diagram of the first embodiment of the character recognition system 100 of the present application.

In this embodiment, the character recognition system 100 includes a series of computer program instructions stored on the memory 11, and when the computer program instructions are executed by the processor 12, the character recognition operations of the embodiments of the present application can be implemented. In some embodiments, the character recognition system 100 may be divided into one or more modules based on the specific operations implemented by the various portions of the computer program instructions. For example, in FIG. 2, the character recognition system 100 may be divided into an acquisition module 101, a processing module 102, a generation module 103, and an output module 104. among them:

The acquisition module 101 is used to acquire character data, and perform image synthesis on each acquired character data with a preset background picture to obtain a character image corresponding to each character data.

Specifically, the character data may be English letters, symbols, numbers, Chinese characters, etc. In this embodiment, the character data includes at least one character. The character data can be captured from the network and then stored in a preset file. When users need to use the character data, they can directly obtain it from the preset file; the character data can also be the characters provided by the business party The data is stored in the preset file. When the user needs to use the character data, it can also be directly obtained from the preset file. Preferably, the preset file is a file in TXT format. A person skilled in the art may obtain the character data in any manner, which will not be repeated here.

The preset background picture is a picture determined by the user according to actual needs. In this embodiment, the preset background picture is preferably a picture grabbed from the Internet with a keyword of "paper". The picture is at least This is a picture, of course, the picture can also be obtained by the user using the camera to shoot various papers. It can be understood that, in other embodiments of the present application, the preset background image may also be a picture of another style, such as a license plate number picture, an ID card picture, and the like.

For example, when the acquired character data has 5 character data, and the preset background picture has 4 images, when image synthesis is performed, preferably, each character data may be separately performed with each background image Image synthesis, so that each character data can synthesize 4 character images, 5 character data can synthesize 20 character images. Of course, when image synthesis is performed, it is not necessary for each character data to be image synthesized with each background image to obtain a character image, which is not limited in this embodiment. In this embodiment, the character data can be combined with multiple background pictures for image synthesis to increase the diversity of character images.

In this embodiment, any existing image synthesis technology may be used to achieve image synthesis. For example, when performing image synthesis, first, the length of the character data, the style of the character data, and the font size of the character data may be used. Determine the length and width of the pixel space occupied by the character data, and after determining the length and width of the pixel space occupied by the character data, select the corresponding pixel area from the pixels of the background picture so that the pixel corresponding to the character data can be inserted into the pixel In the area, and replace the pixels originally in the pixel area. It can be understood that, in other embodiments, instead of pixel replacement, the pixel superposition method may be used directly, that is, each pixel corresponding to the character data is respectively performed with each pixel corresponding to the pixel area Superposition, the superimposed pixel value is used as the pixel value of each pixel in the pixel area.

The processing module 102 is used to perform random disturbance processing on the synthesized character image to obtain different types of character images.

Specifically, the random disturbance processing includes Gaussian blur processing, Gaussian noise processing, small-scale rotation processing of the picture, and contrast processing and color change processing of the picture. Among them, the Gaussian blur processing of the picture refers to the Gaussian filtering of the picture with a certain mean and variance; the Gaussian noise processing of the picture refers to adding Gaussian noise to the three color channels of the picture. Unlike Gaussian blur, This is directly superimposed on the value, and Gaussian blur is to filter the picture; small-scale rotation of the picture refers to determining the center point to be rotated according to the field frame, or you can directly take the center of the picture as the center point of rotation, This can be adjusted according to the actual business needs, and then rotated by an angle according to the center point; the contrast processing of the picture refers to the random change of the S (Saturation) and V (Value lightness) of the picture in the HSV color space; the picture The color change process refers to randomly changing the H (Hue hue) of the picture in the HSV color space.

In this embodiment, different types of character images can be obtained by using at least one of the above-mentioned perturbation processing methods on the synthesized image, for example, a character image with a rotated pattern, a noisy character image, and an inclined character can be obtained Images etc. By disturbing the synthesized image, the diversity of character images can be increased, so that the data of the training sample is more abundant, so that the character recognition model trained by the training sample can have higher recognition accuracy.

The generating module 103 is used to input the different types of character images into a deep learning network for training to generate a character recognition model.

Specifically, before inputting different types of character images into the deep learning network, the character image needs to be pre-processed to convert the character image into a required feature vector, and then the required feature vector is input into the deep learning network For training.

In this embodiment, the deep learning network is preferably a CRNN model. The CRNN model is a joint model of a convolutional neural network and a recurrent neural network. The CRNN model is an end-to-end trainable model, which has the following Advantages: 1) The input data can be of any length (the image width is arbitrary, the word length is arbitrary); 2) The training set does not require character calibration; 3) Both the dictionary with and without the dictionary (sample) can be used; 4) Performance Good, and the model is small (fewer parameters).

In a specific embodiment, the CRNN model includes a VGG16 layer, two long and short-term memory network LSTM layers, and two fully connected FC layers, where the VGG16 layer is composed of 13 convolutional layers and 3 fully linked layers For extracting the spatial features of character images; the two long and short-term memory network LSTM layers are used to extract the temporal features of character images to obtain the contextual relationship of the text to be trained and recognized; the two are fully connected The FC layer is used to classify the extracted spatial and temporal features. Compared with the existing CRNN model, the CRNN model in this embodiment adds a fully connected FC layer to speed up the convergence of training.

The output module 104 is used to input the character image to be recognized into the character recognition model, and output the recognition result of the character image to be recognized.

In this embodiment, when a user needs to recognize a character, he only needs to collect a character image of the character to be recognized, and then input the character image into a character recognition model, and the character recognition model can recognize the character corresponding to the character image. In this embodiment, the character recognition model may be stored in a local character recognition terminal or may be stored in a server, which is specifically selected according to the actual needs of the user, and is not limited in this embodiment.

Through the above program modules 101-104, the character recognition system 100 proposed in this application acquires character data, and synthesizes each acquired character data with a preset background picture to obtain a character image corresponding to each character data; Perform random perturbation processing on the synthesized character images to obtain different types of character images; input the different types of character images into a deep learning network for training to generate a character recognition model; and input character images to be recognized into the characters In the recognition model, the recognition result of the character image to be recognized is output. In this way, a variety of training sample data can be generated as needed to solve the problem of small character recognition range and low accuracy due to uneven distribution of real data of training samples in the prior art, increasing the character recognition range and increasing character recognition accuracy .

Referring to FIG. 3, it is a program module diagram of the second embodiment of the character recognition system 100 of the present application. In this embodiment, the character recognition system 100 includes a series of computer program instructions stored on the memory 11, and when the computer program instructions are executed by the processor 12, the character recognition operations of the embodiments of the present application can be implemented. In some embodiments, the character recognition system 100 may be divided into one or more modules based on the specific operations implemented by the various parts of the computer program instructions. For example, in FIG. 3, the character recognition system 100 may be divided into an acquisition module 101, a processing module 102, a generation module 103, an output module 104, a test module 105, and an adjustment module 106. The program modules 101-104 are the same as the first embodiment of the character recognition system 100 of the present application, and on this basis, a test module 105 and an adjustment module 106 are added. among them:

In this embodiment, when a user needs to recognize a character, he only needs to collect a character image of the character to be recognized, and then input the character image into a character recognition model, and the character recognition model can recognize the character corresponding to the character image. In this embodiment, the character recognition model may be stored in a local character recognition terminal or may be stored in a server, which is selected according to the actual needs of the user.

The testing module 105 is used to test the character recognition accuracy of the character recognition model.

Specifically, after the character recognition model is generated, the recognition accuracy of the character recognition model on real character image data needs to be tested.

In one embodiment, the user inputs character images of several real characters into the character recognition model, outputs the recognition result corresponding to the real character, and then calculates the accuracy rate of character recognition according to the output recognition result. It can be understood that, in order to obtain accurate calculation results of the character recognition rate, the amount of real character data input into the character recognition model should be as much as possible.

When calculating the accuracy of character recognition, the recognition result output by the character recognition model can be compared with the pre-stored character data to determine whether the character recognition model is true or not for the character recognition. If the data recognition is correct, you can count and accumulate 1 until after all character recognition is completed, divide the calculated accumulated value by the number of characters input into the character recognition model to obtain the recognition accuracy rate of the character recognition model for real character image data .

The adjustment module 106 is used to adjust the character recognition model if the recognition accuracy rate is lower than a preset threshold.

Specifically, after the character recognition accuracy rate of the character recognition model is obtained, the character recognition accuracy rate is compared with a preset threshold, and if the character recognition accuracy rate is lower than the preset threshold, the character Identify the model and make adjustments. In this embodiment, the preset threshold is the lowest value of the character recognition accuracy rate set in advance, for example, the preset threshold is 90%. The preset threshold can be set according to the actual needs of the user, and the preset threshold after the setting can be further modified according to the actual needs.

It should be noted that when the character recognition model is adjusted in this embodiment, the character recognition model is only fine-tuned, and does not need to be adjusted too much.

Specifically, the step of adjusting the character recognition model includes:

Step A. Freeze the parameters of the VGG16 layer.

In this embodiment, when the character recognition model is adjusted, the parameters of the VGG16 layer are not changed, that is, the parameters of the VGG16 layer are frozen to prevent the adjustment of the character recognition model. The parameters of the VGG16 layer are adjusted under the stimulation of the training sample data.

Step B: Adjust the parameters of the two long-short-term memory network LSTM layers and the two fully connected FC layers.

In this embodiment, when the character recognition model is adjusted, the parameters of the two long-short-term memory network LSTM layers and the two fully connected FC layers are adjusted, specifically, by releasing the two The parameters of the long- and short-term memory network LSTM layer and two fully connected FC layers, and the learning rate is set to decay every several epochs until it decay to a boundary value.

Step C: Train the adjusted character recognition model using real character image data.

In this embodiment, while adjusting the parameters of the two long-short-term memory network LSTM layers and the two fully connected FC layers, the real character image is input to the character recognition after the adjustment parameters are entered The model is further trained to obtain an adjusted character recognition model. After obtaining the adjusted character recognition model, the test module 105 is used to test the recognition accuracy of the model. If the test result meets the requirements, the character recognition model training is completed; if the test module 105 is used to recognize the character When the test result obtained by the model test still does not meet the requirements, step A to step C are repeated until the recognition accuracy of the obtained character recognition model reaches the requirements.

Through the above program modules 101-106, the character recognition system 100 proposed in this application acquires character data, and synthesizes each acquired character data with a preset background picture to obtain a character image corresponding to each character data; Perform random perturbation processing on the synthesized character images to obtain different types of character images; input the different types of character images into a deep learning network for training to generate a character recognition model; input character images to be recognized into the character recognition In the model, the recognition result of the character image to be recognized is output; the character recognition accuracy rate of the character recognition model is tested; and if the recognition accuracy rate is lower than a preset threshold, the character recognition model is adjusted. In this way, by fine-tuning the character recognition model when the character recognition model does not reach the preset recognition accuracy, the accuracy of character recognition is improved.

In addition, this application also proposes a character recognition method.

Referring to FIG. 4, it is a schematic diagram of an implementation process of the first embodiment of the character recognition method of the present application. In this embodiment, according to different requirements, the execution order of the steps in the flowchart shown in FIG. 4 may be changed, and some steps may be omitted.

In step S500, character data is acquired, and each acquired character data is image synthesized with a preset background picture to obtain a character image corresponding to each character data.

For example, when the acquired character data has 5 character data, and the preset background picture has 4 images, when image synthesis is performed, preferably, each character data may be separately performed with each background image Image synthesis, so that each character data can synthesize 4 character images, 5 character data can synthesize 20 character images. Of course, when image synthesis is performed, it is not necessary for each character data to be image synthesized with each background image to obtain a character image, which is not limited in this embodiment. In this embodiment, the character data can be increased in versatility by performing image synthesis with character data and multiple background pictures.

Step S502: Perform random perturbation processing on the synthesized character image to obtain character images of different types.

Step S504, input the character images of different types into the deep learning network for training to generate a character recognition model.

Step S506: Input the character image to be recognized into the character recognition model, and output the recognition result of the character image to be recognized.

Through the above steps S500-S506, the character recognition method proposed in this application acquires character data, and performs image synthesis on each acquired character data with a preset background picture to obtain a character image corresponding to each character data; Random image perturbation processing to obtain different types of character images; input the different types of character images into a deep learning network for training to generate a character recognition model; and input character images to be recognized into the character recognition model , Output the recognition result of the character image to be recognized. In this way, a variety of training sample data can be generated as needed to solve the problem of small character recognition range and low accuracy due to uneven distribution of real data of training samples in the prior art, increasing the character recognition range and increasing character recognition accuracy .

Referring to FIG. 5, it is a schematic diagram of an implementation process of the second embodiment of the character recognition method of the present application. In this embodiment, according to different requirements, the execution order of the steps in the flowchart shown in FIG. 5 may be changed, and some steps may be omitted.

In step S600, character data is acquired, and each acquired character data is image synthesized with a preset background picture to obtain a character image corresponding to each character data.

Step S602, random disturbance processing is performed on the synthesized character image to obtain character images of different types.

Step S604, input the character images of different types into the deep learning network for training to generate a character recognition model.

Step S606: Input the character image to be recognized into the character recognition model, and output the recognition result of the character image to be recognized.

The above steps S600-S606 are similar to the steps S500-S506, and will not be repeated in this embodiment.

Step S608: Test the character recognition accuracy of the character recognition model.

Step S610: If the recognition accuracy rate is lower than a preset threshold, adjust the character recognition model.

Specifically, the step of adjusting the character recognition model includes:

Step A. Freeze the parameters of the VGG16 layer.

Through the above steps S600-S610, the character recognition method proposed in this application acquires character data, and performs image synthesis on each acquired character data with a preset background picture to obtain a character image corresponding to each character data; Random image perturbation processing to obtain different types of character images; input the different types of character images into the deep learning network for training to generate a character recognition model; input character images to be recognized into the character recognition model , Output the recognition result of the character image to be recognized; test the character recognition accuracy rate of the character recognition model; and if the recognition accuracy rate is lower than a preset threshold, adjust the character recognition model. In this way, by fine-tuning the character recognition model when the character recognition model does not reach the preset recognition accuracy, the accuracy of character recognition is improved.

The sequence numbers of the above embodiments of the present application are for description only, and do not represent the advantages and disadvantages of the embodiments.

Through the description of the above embodiments, those skilled in the art can clearly understand that the methods in the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware, but in many cases the former is better Implementation. Based on this understanding, the technical solutions of the present application can be embodied in the form of software products in essence or part of contributions to the existing technology, and the computer software products are stored in a storage medium (such as ROM / RAM, magnetic disk, The CD-ROM includes several instructions to enable a server (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the embodiments of the present application.

The above are only the preferred embodiments of the present application, and do not limit the scope of the patent of the present application. Any equivalent structure or equivalent process transformation made by the description and drawings of this application, or directly or indirectly used in other related technical fields The same reason is included in the patent protection scope of this application.

Claims

A character recognition method applied to a server, characterized in that the method includes:

Obtain character data, and synthesize each acquired character data with a preset background picture to obtain a character image corresponding to each character data;

Randomly disturb the synthesized character image to obtain different types of character images;

Input the different types of character images into the deep learning network for training to generate a character recognition model; and

The character image to be recognized is input into the character recognition model, and the recognition result of the character image to be recognized is output.
The character recognition method according to claim 1, wherein the deep learning network is a CRNN model, and the CRNN model includes a VGG16 layer, two long and short-term memory network LSTM layers, and two fully connected FC layers, wherein , The VGG16 layer is used to extract the spatial features of character images, the two long and short-term memory network LSTM layers are used to extract the timing features of character images, and the two fully connected FC layers are used to extract Classify the spatial and temporal features of
The character recognition method according to claim 2, wherein after the step of inputting the character images of different types into a deep learning network for training to generate a character recognition model, the method further comprises:

Testing the character recognition accuracy of the character recognition model; and

If the recognition accuracy rate is lower than a preset threshold, the character recognition model is adjusted.
The character recognition method according to claim 3, wherein the step of adjusting the character recognition model includes:

Freeze the parameters of the VGG16 layer;

Adjusting the parameters of the two long-short-term memory network LSTM layers and the two fully connected FC layers; and

The real character image data is used to train the adjusted character recognition model.
The character recognition method according to claim 1, wherein the random disturbance processing includes: Gaussian blur processing, Gaussian noise processing, small-scale rotation processing of the picture, contrast change processing of the picture, and color change processing of the picture At least one.
The character recognition method according to claim 2, wherein the random disturbance processing includes: Gaussian blur processing, Gaussian noise processing, small-scale rotation processing of the picture, contrast change processing of the picture, and color change processing of the picture At least one.
The character recognition method according to claim 3, wherein the random disturbance processing includes: Gaussian blur processing, Gaussian noise processing, small-scale rotation processing of the picture, contrast change processing of the picture, and color change processing of the picture At least one.
The character recognition method according to claim 4, wherein the random disturbance processing includes: Gaussian blur processing, Gaussian noise processing, small-scale rotation processing of the picture, contrast change processing of the picture, and color change processing of the picture At least one.
A server, characterized in that the server includes a memory and a processor, and a character recognition system that can run on the processor is stored on the memory, and the character recognition system is implemented as follows when executed by the processor step:

Obtain character data, and synthesize each acquired character data with a preset background picture to obtain a character image corresponding to each character data;

Randomly disturb the synthesized character image to obtain different types of character images;

Input the different types of character images into the deep learning network for training to generate a character recognition model; and

The character image to be recognized is input into the character recognition model, and the recognition result of the character image to be recognized is output.
The server according to claim 9, wherein the deep learning network is a CRNN model, and the CRNN model includes a VGG16 layer, two long and short-term memory network LSTM layers, and two fully connected FC layers. The VGG16 layer is used to extract the spatial features of the character image, the two long and short-term memory network LSTM layers are used to extract the timing features of the character image, and the two fully connected FC layers are used to extract the space Features and timing features are classified.
The server according to claim 10, wherein when the character recognition system is executed by the processor, the following steps are further implemented:

Testing the character recognition accuracy of the character recognition model; and

If the recognition accuracy rate is lower than a preset threshold, the character recognition model is adjusted.
The server according to claim 11, wherein the step of adjusting the character recognition model includes:

Freeze the parameters of the VGG16 layer;

Adjusting the parameters of the two long-short-term memory network LSTM layers and the two fully connected FC layers; and

The real character image data is used to train the adjusted character recognition model.
The server according to claim 9, wherein the random disturbance processing includes at least one of Gaussian blur processing, Gaussian noise processing, small-scale rotation processing of the image, contrast change processing of the image, and color change processing of the image Species.
The server according to claim 10, wherein the random disturbance processing includes at least one of Gaussian blur processing, Gaussian noise processing, small-scale rotation processing of the image, contrast change processing of the image, and color change processing of the image Species.
The server according to claim 11, wherein the random disturbance processing includes at least one of Gaussian blur processing, Gaussian noise processing, small-scale rotation processing of the image, contrast change processing of the image, and color change processing of the image Species.
The server according to claim 12, wherein the random disturbance processing includes at least one of Gaussian blur processing, Gaussian noise processing, small-scale rotation processing of the image, contrast change processing of the image, and color change processing of the image Species.
A computer-readable storage medium storing a character recognition system, the character recognition system may be executed by at least one processor, so that the at least one processor performs the following steps:

Obtain character data, and synthesize each acquired character data with a preset background picture to obtain a character image corresponding to each character data;

Randomly disturb the synthesized character image to obtain different types of character images;

Input the different types of character images into the deep learning network for training to generate a character recognition model; and

The character image to be recognized is input into the character recognition model, and the recognition result of the character image to be recognized is output.
The computer-readable storage medium of claim 17, wherein the deep learning network is a CRNN model, and the CRNN model includes a VGG16 layer, two long and short-term memory network LSTM layers, and two fully connected FC layers , Where the VGG16 layer is used to extract the spatial features of the character image, the two long and short-term memory network LSTM layers are used to extract the timing features of the character image, and the two fully connected FC layers are used to The extracted spatial features and time series features are classified.
The computer-readable storage medium of claim 18, wherein after the step of inputting the character images of different types into a deep learning network for training to generate a character recognition model, further comprising:

Testing the character recognition accuracy of the character recognition model; and

If the recognition accuracy rate is lower than a preset threshold, the character recognition model is adjusted.
The computer-readable storage medium of claim 19, wherein the step of adjusting the character recognition model includes:

Freeze the parameters of the VGG16 layer;

Adjusting the parameters of the two long-short-term memory network LSTM layers and the two fully connected FC layers; and

The real character image data is used to train the adjusted character recognition model.