A training method and system for data recognition model

Abstract

The present invention is applicable to the field of data recognition technology, and in particular, relates to a training method and system for a data recognition model, the method comprising: obtaining a fixed training data set; comparing all characters contained in the fixed training data set to obtain a classified character group; counting the number of strokes of the characters in each classified character group, and sorting the characters in the classified character group; deforming all characters in the classified character group to obtain a deformed character data set; integrating the deformed character data set and the fixed training data set to obtain an auxiliary training data set, and training the data recognition model. The present invention generates additional auxiliary training data by auxiliary expansion of the original training data, and the auxiliary training data can cover data similar to the original training data, so as to achieve the purpose of comprehensive training of the data recognition model, so that the data recognition model can have higher recognition accuracy during the data recognition process.

Description

Technical Field

[0001] The present invention belongs to the technical field of data recognition, and in particular relates to a training method and system for a data recognition model. Background Art

[0002] With the development of network technology, data identification has become an indispensable technical means to realize different computer functions. The so-called data identification is to use computer technology to analyze the data to be identified, so as to determine the content of the data.

[0003] In the current data recognition model, there is text recognition, which analyzes the digitized image to determine the text information contained in the digitized image. Therefore, with the continuous development of data recognition technology, the accuracy of text recognition is getting higher and higher.

[0004] However, before the data recognition model is applied, it needs to be trained. The existing training method mainly selects data from a fixed training data set for the data recognition model to recognize, and performs real-time correction based on the recognition results to finally obtain a trained data recognition model. The data recognition model obtained in this way has high recognition accuracy for data that is the same as the training data, but once the data to be recognized is different from the training data, its recognition degree will be greatly reduced. Summary of the invention

[0005] The purpose of the embodiment of the present invention is to provide a training method for a data recognition model, aiming to solve the problem raised in the third part of the background technology.

[0006] The embodiment of the present invention is implemented as follows: a method for training a data recognition model, the method comprising:

[0007] Obtain a fixed training data set, wherein the fixed training data set contains the same font information of all characters;

[0008] Comparing all characters included in the fixed training data set with each other to obtain a classified character group, wherein the classified character group includes at least one character;

[0009] Count the number of strokes of each character in the classified character group, and sort the characters in the classified character group according to the number of strokes;

[0010] Performing deformation processing on all characters in the classified character group to obtain a deformed character data set, wherein the deformed character data set contains a correspondence between each character and the deformed character;

[0011] The deformed text dataset and the fixed training dataset are integrated to obtain an auxiliary training dataset, and the auxiliary training dataset is used to train the data recognition model.

[0012] Preferably, the step of comparing all the characters contained in the fixed training data set to obtain the classified character group specifically includes:

[0013] Counting the number of strokes of all characters contained in the fixed training data set, and classifying the characters according to the number of strokes to obtain stroke-classified character groups, wherein the strokes of the characters in the stroke-classified character groups are the same;

[0014] Numbering the stroke classification character groups according to the number of strokes, selecting a character in a stroke classification character group each time, and comparing it with other characters in the stroke classification character group to determine distinguishing stroke features;

[0015] The characters whose number of stroke distinguishing features is less than a preset value are classified into one category to obtain a classified character group.

[0016] Preferably, the step of counting the number of strokes of each character in the classified character group and sorting the characters in the classified character group according to the number of strokes specifically includes:

[0017] Count the number of strokes of each character in each classified character group;

[0018] Sort the characters in the classified character group according to the number of strokes.

[0019] Preferably, the step of deforming all characters in the classified character group to obtain a deformed character data set specifically includes:

[0020] Each time, a text deformation scheme is read from the preset deformation database;

[0021] According to the text deformation scheme, all the characters in the classified text group are deformed accordingly to obtain a deformed text group;

[0022] Repeat the above steps until all text deformation schemes in the preset deformation database are read to generate a deformed text data set.

[0023] Preferably, the step of training the data recognition model using the auxiliary training data set specifically includes:

[0024] Each time, read a word from a fixed training data set for training;

[0025] The deformed text in the deformed text dataset corresponding to the text is read for training.

[0026] Preferably, the step of deforming all characters in the classified character group at least includes deforming the characters in terms of aspect ratio.

[0027] Preferably, when sorting the characters in the classified character group according to the number of strokes, ascending order is used for sorting.

[0028] Another object of an embodiment of the present invention is to provide a data recognition model training system, the system comprising:

[0029] A data acquisition module, used to acquire a fixed training data set, wherein the fixed training data set contains the same font information of all characters;

[0030] A text comparison module, used to compare all the texts contained in the fixed training data set to obtain a classified text group, wherein the classified text group contains at least one text;

[0031] The stroke counting module is used to count the number of strokes of each character in the classified character group and sort the characters in the classified character group according to the number of strokes;

[0032] A text deformation module, used for deforming all the characters in the classified text group to obtain a deformed text data set, wherein the deformed text data set contains a correspondence between each character and the deformed text;

[0033] The data integration module is used to integrate the deformed text data set and the fixed training data set to obtain an auxiliary training data set, and use the auxiliary training data set to train the data recognition model.

[0034] Preferably, the text comparison module includes:

[0035] A data statistics unit, used for counting the number of strokes of all characters contained in the fixed training data set, and classifying the characters according to the number of strokes to obtain stroke classification character groups, wherein the strokes of the characters in the stroke classification character groups are the same;

[0036] A feature comparison unit is used to number the stroke classification character groups according to the number of strokes, select a character in a stroke classification character group each time, compare it with other characters in the stroke classification character group, and determine distinguishing stroke features;

[0037] The data classification unit is used to classify characters whose number of stroke distinguishing features is less than a preset value into one category to obtain a classified character group.

[0038] Preferably, the stroke statistics module includes:

[0039] A stroke analysis unit, used to count the number of strokes of each character in each classified character group;

[0040] The stroke sorting unit is used to sort the characters in the classified character group according to the number of strokes.

[0041] A data recognition model training method provided by an embodiment of the present invention generates additional auxiliary training data by auxiliary expansion of the original training data. The auxiliary training data can cover data similar to the original training data to achieve the purpose of comprehensive training of the data recognition model, so that the data recognition model can have higher recognition accuracy during the data recognition process. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flowchart of a method for training a data recognition model provided by an embodiment of the present invention;

[0043] Figure 2 A flowchart of the steps of comparing all the characters contained in a fixed training data set to obtain a classified character group provided by an embodiment of the present invention;

[0044] Figure 3 A flowchart of the steps of counting the number of strokes of characters in each classified character group and sorting the characters in the classified character group according to the number of strokes provided by an embodiment of the present invention;

[0045] Figure 4 A flowchart of the steps of deforming all characters in a classified character group to obtain a deformed character data set provided by an embodiment of the present invention;

[0046] Figure 5 A flowchart of the steps of training a data recognition model using an auxiliary training data set provided in an embodiment of the present invention;

[0047] Figure 6 An architectural diagram of a data recognition model training system provided by an embodiment of the present invention;

[0048] Figure 7 An architectural diagram of a text comparison module provided in an embodiment of the present invention;

[0049] Figure 8 This is an architecture diagram of a stroke statistics module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below 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 invention and are not intended to limit the present invention.

[0051] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0052] Among the current data recognition models, there are text recognition models. It analyzes digitized images to determine the text information contained in the digitized images. Therefore, with the continuous development of data recognition technology, the accuracy of text recognition is getting higher and higher. However, before the data recognition model is applied, it needs to be trained. The existing training method mainly selects data from a fixed training data set for the data recognition model to recognize, and performs real-time correction based on the recognition results, and finally obtains a trained data recognition model. The data recognition model obtained in this way has high recognition accuracy for data that is the same as the training data, but once the data to be recognized is different from the training data, its recognition degree will be greatly reduced.

[0053] like Figure 1 The above is a flowchart of a method for training a data recognition model provided by an embodiment of the present invention, and the method includes:

[0054] S100, obtaining a fixed training data set, wherein the fixed training data set contains the same font information of all characters.

[0055] In this step, a fixed training data set is obtained, wherein the fixed training data set includes all texts in a font. For example, taking bold as an example, the fixed training data set includes all bold texts.

[0056] S200, comparing all characters included in the fixed training data set to obtain a classified character group, wherein the classified character group includes at least one character.

[0057] In this step, all the characters contained in the fixed training data set are compared with each other. Since the number of characters is very large, there are about 7,000 commonly used characters, and there are more than 90,000 Chinese characters. Therefore, if a one-by-one comparison method is adopted, a huge amount of data processing will be generated, affecting the data processing efficiency. In order to solve the above problem, the characters can be classified and similar characters can be compared, which can greatly reduce the amount of data comparison. After classification in the above method, a classified character group is obtained, and the classified character group contains at least one character.

[0058] S300, counting the number of strokes of the characters in each classified character group, and sorting the characters in the classified character group according to the number of strokes.

[0059] In this step, the number of strokes of the characters in each classified character group is counted. After the preliminary classification, it is only necessary to count the number of strokes of the characters in the same classified character group, and then sort the characters in the classified character group according to the number of strokes. This is because when performing character recognition, if a one-to-one comparison method is adopted, multiple comparisons are required, which slows down the recognition result. In the present invention, the character to be recognized can be compared with the character with the smallest number of strokes in the classified character group. When the two are consistent, the character to be recognized can be compared one by one with other characters in the corresponding classified character group, thereby finally determining the recognition result of the character to be recognized.

[0060] S400, deforming all characters in the classified character group to obtain a deformed character data set, wherein the deformed character data set contains a correspondence between each character and the deformed character.

[0061] In this step, all the characters in the classified character group are deformed. For the characters to be recognized, due to the display device or other factors, the characters to be recognized often appear deformed to a certain extent. Therefore, only by comparing the characters to be recognized with the standard font, inaccurate recognition is likely to occur. Therefore, by deforming the standard font, the data recognition model can recognize the characters with slight deformation to ensure the recognition accuracy. The deformed character data set contains the correspondence between each character and the deformed character.

[0062] S500, integrating the deformed text data set and the fixed training data set to obtain an auxiliary training data set, and using the auxiliary training data set to train the data recognition model.

[0063] In this step, the deformed text data set and the fixed training data set are integrated to obtain an auxiliary training data set. The auxiliary training data set records the standard font and the deformed font. A mapping relationship is established between the standard font and the deformed font, which can assist the data recognition model in training and determine the standard font corresponding to the deformed text to improve the accuracy of text recognition.

[0064] like Figure 2 As shown, as a preferred embodiment of the present invention, the step of comparing all the characters contained in the fixed training data set to obtain the classified character group specifically includes:

[0065] S201. Count the number of strokes of all the characters in the fixed training dataset, and classify the characters according to the number of strokes to obtain a group of stroke-classified characters, where the characters within the group have the same number of strokes.

[0066] In this step, count the number of strokes of all the characters in the fixed training dataset. Each character is composed of individual strokes, and for similar characters, the number of strokes is necessarily similar. Therefore, by counting the number of strokes of the characters, the characters can be classified according to the number of strokes to obtain a group of stroke-classified characters, where the characters within the group have the same number of strokes, which facilitates comparison.

[0067] S202. Number the groups of stroke-classified characters according to the number of strokes. Each time, select one character within a group of stroke-classified characters and compare it with the other characters within the same group of stroke-classified characters to determine the distinguishing stroke features.

[0068] In this step, number the groups of stroke-classified characters according to the number of strokes. For example, divide the groups of stroke-classified characters according to the number of strokes to obtain the first group of stroke-classified characters, the second group of stroke-classified characters, etc. Each time, select one character within a group of stroke-classified characters. Based on this character, compare it with the other characters within the same group of stroke-classified characters, and use the comparison method to determine the difference points between the two characters being compared, that is, determine the distinguishing stroke features. For example, for the characters '干' and '于', the distinguishing stroke feature between them is the hooked part. Therefore, the character '于' can be regarded as adding the distinguishing stroke feature to the character '干'. Then, during the character recognition process, first compare the character to be recognized with the character '干' to initially determine that the character to be recognized is '干', and then recognize whether it has a hooked part to further determine whether the character to be recognized is '于'.

[0069] S203. Classify the characters with the number of distinguishing stroke features less than the preset value into one category to obtain a group of classified characters.

[0070] In this step, classify the characters with the number of distinguishing stroke features less than the preset value into one category to obtain a group of classified characters. For example, if the preset value is 3, then classify the characters with the number of distinguishing stroke features less than 3 into one category to obtain a group of classified characters.

[0071] As Figure 3 shown, as a preferred embodiment of the present invention, the step of counting the number of strokes of each character within each group of classified characters and sorting the characters within the group of classified characters according to the number of strokes specifically includes:

[0072] S301. Count the number of strokes of each character within each group of classified characters.

[0073] S302, sorting the characters in the classified character group according to the number of strokes.

[0074] In this step, the number of strokes of each character in each classified character group is counted again. This is because within each classified character group, if no sorting is performed, confusion is likely to occur during comparison. Therefore, the characters in the classified character group need to be sorted according to the number of strokes. When sorting the characters in the classified character group according to the number of strokes, ascending order is used.

[0075] like Figure 4 As shown, as a preferred embodiment of the present invention, the step of deforming all characters in the classified character group to obtain a deformed character data set specifically includes:

[0076] S401, reading a text deformation scheme from a preset deformation database each time.

[0077] S402, performing corresponding deformation processing on all characters in the classified character group according to the character deformation scheme to obtain a deformed character group.

[0078] In this step, in order to ensure the stability of data processing, a text deformation scheme can be read from the preset deformation database each time. Each text deformation scheme records the specific deformation method, such as widening the font, lengthening the font, or tilting the font. According to the text deformation scheme, all the text in the classified text group is deformed accordingly to obtain a deformed text group; the step of deforming all the text in the classified text group at least includes deforming the aspect ratio of the text.

[0079] S403, repeating the above steps until all text deformation schemes in the preset deformation database are read to generate a deformed text data set.

[0080] like Figure 5 As shown, as a preferred embodiment of the present invention, the step of training the data recognition model using the auxiliary training data set specifically includes:

[0081] S501, reading one character in a fixed training data set each time for training.

[0082] S502, reading the deformed characters in the deformed character data set corresponding to the character for training.

[0083] In this step, each time a character in a fixed training data set is read for training, the data recognition model can recognize the standard font, so that the recognition accuracy of the standard font is guaranteed. Then the deformed character in the deformed character data set corresponding to the character is read for training. At this time, the deformed character that is slightly deformed from the standard font can also be recognized, thereby ensuring the recognition accuracy of the data recognition model.

[0084] like Figure 6 As shown, a data recognition model training system provided by the present invention comprises:

[0085] The data acquisition module 100 is used to acquire a fixed training data set, wherein the fixed training data set contains the same font information of all characters.

[0086] In the present system, the data acquisition module 100 acquires a fixed training data set, wherein the fixed training data set includes all texts in a font.

[0087] The text comparison module 200 is used to compare all the texts contained in the fixed training data set to obtain a classified text group, wherein the classified text group contains at least one text.

[0088] In this system, the text comparison module 200 compares all the texts contained in the fixed training data set. Since the number of texts is very large, if a one-by-one comparison method is adopted, a huge amount of data processing will be generated, affecting the data processing efficiency. In order to solve the above problem, the texts can be classified and similar texts can be compared, thereby greatly reducing the amount of data comparison. After classification in the above manner, a classified text group is obtained, and the classified text group contains at least one text.

[0089] The stroke counting module 300 is used to count the number of strokes of each character in the classified character group and sort the characters in the classified character group according to the number of strokes.

[0090] In this system, the stroke counting module 300 counts the number of strokes of each character in the classified character group. After the preliminary classification, it is only necessary to count the number of strokes of the characters in the same classified character group, and then sort the characters in the classified character group according to the number of strokes. This is because when performing character recognition, if a one-to-one comparison method is adopted, multiple comparisons are required, which slows down the recognition result. In the present invention, the character to be recognized can be compared with the character with the smallest number of strokes in the classified character group. When the two are consistent, the character to be recognized can be compared one by one with other characters in the corresponding classified character group, so as to finally determine the recognition result of the character to be recognized.

[0091] The text deformation module 400 is used to deform all the characters in the classified text group to obtain a deformed text data set, wherein the deformed text data set contains the correspondence between each character and the deformed text.

[0092] In this system, the text deformation module 400 deforms all the texts in the classified text group. For the text to be recognized, due to the display device or other factors, the text to be recognized often undergoes a certain degree of deformation. Therefore, only by comparing the text to be recognized with the standard font, inaccurate recognition is likely to occur. Therefore, by deforming the standard font, the data recognition model can recognize the text with slight deformation to ensure the recognition accuracy. The deformed text data set contains the correspondence between each character and the deformed text.

[0093] The data integration module 500 is used to integrate the deformed text data set and the fixed training data set to obtain an auxiliary training data set, and use the auxiliary training data set to train the data recognition model.

[0094] In this system, the data integration module 500 integrates the deformed text data set and the fixed training data set to obtain an auxiliary training data set. The auxiliary training data set records the standard font and the deformed font. A mapping relationship is established between the standard font and the deformed font, so as to assist the data recognition model in training and determine the standard font corresponding to the deformed text to improve the accuracy of text recognition.

[0095] like Figure 7 As shown, as a preferred embodiment of the present invention, the text comparison module 200 includes:

[0096] The data statistics unit 201 is used to count the number of strokes of all characters included in the fixed training data set, and classify the characters according to the number of strokes to obtain stroke classification character groups, where the strokes of the characters in the stroke classification character groups are the same.

[0097] In this module, the data statistics unit 201 counts the number of strokes of all the characters contained in the fixed training data set. Each character is composed of a single stroke, and similar characters must have similar numbers of strokes. Therefore, by counting the number of strokes of the characters, the characters can be classified according to the number of strokes to obtain stroke-classified character groups. The strokes of the characters in the stroke-classified character groups are the same, which makes it convenient to compare.

[0098] The feature comparison unit 202 is used to number the stroke classification character groups according to the number of strokes, select a character in a stroke classification character group each time, compare it with other characters in the stroke classification character group, and determine distinguishing stroke features.

[0099] In this module, the feature comparison unit 202 numbers the stroke classification character groups according to the number of strokes, for example, divides a plurality of stroke classification character groups according to the number of strokes to obtain a first stroke classification character group, a second stroke classification character group, etc. Each time, a character in a stroke classification character group is selected, and based on the character, it is compared with other characters in the stroke classification character group, and the difference between the two characters to be compared is determined by comparison, that is, the distinguishing stroke features are determined.

[0100] The data classification unit 203 is used to classify characters whose number of stroke distinguishing features is less than a preset value into one category to obtain a classified character group.

[0101] In this module, the data classification unit 203 classifies characters whose number of stroke distinguishing features is less than a preset value into one category to obtain a classified character group.

[0102] like Figure 8 As shown, as a preferred embodiment of the present invention, the stroke statistics module 300 includes:

[0103] The stroke analysis unit 301 is used to count the number of strokes of each character in each classified character group.

[0104] The stroke sorting unit 302 is used to sort the characters in the classified character group according to the number of strokes.

[0105] In this module, the number of strokes of each character in each classified character group is counted again. This is because within each classified character group, if the characters are not sorted, confusion will easily occur during comparison. Therefore, the characters in the classified character group need to be sorted according to the number of strokes.

[0106] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0107] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0108] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0109] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for training a data recognition model, characterized in that: The method comprises: Obtain a fixed training data set, wherein the fixed training data set contains the same font information of all characters; Comparing all characters included in the fixed training data set with each other to obtain a classified character group, wherein the classified character group includes at least one character; Count the number of strokes of each character in the classified character group, and sort the characters in the classified character group according to the number of strokes; Performing deformation processing on all characters in the classified character group to obtain a deformed character data set, wherein the deformed character data set contains a correspondence between each character and the deformed character; The deformed text dataset and the fixed training dataset are integrated to obtain an auxiliary training dataset, and the data recognition model is trained using the auxiliary training dataset; The step of comparing all the characters contained in the fixed training data set to obtain the classified character group specifically includes: Counting the number of strokes of all characters contained in the fixed training data set, and classifying the characters according to the number of strokes to obtain stroke-classified character groups, wherein the strokes of the characters in the stroke-classified character groups are the same; Numbering the stroke classification character groups according to the number of strokes, selecting a character in a stroke classification character group each time, and comparing it with other characters in the stroke classification character group to determine distinguishing stroke features; The characters whose number of stroke distinguishing features is less than a preset value are classified into one category to obtain a classified character group.

2. The data recognition model training method according to claim 1, characterized in that: The step of deforming all characters in the classified character group to obtain a deformed character data set specifically includes: Each time, a text deformation scheme is read from the preset deformation database; According to the text deformation scheme, all the characters in the classified text group are deformed accordingly to obtain a deformed text group; Repeat the above steps until all text deformation schemes in the preset deformation database are read to generate a deformed text data set.

3. The training method of the data recognition model according to claim 1, characterized in that: The step of training the data recognition model using the auxiliary training data set specifically includes: Each time, read a word from a fixed training data set for training; The deformed text in the deformed text dataset corresponding to the text is read for training.

4. The method for training a data recognition model according to claim 1, characterized in that: The step of deforming all characters in the classified character group at least includes deforming the characters in terms of aspect ratio.

5. The method for training a data recognition model according to claim 1, characterized in that: When sorting the characters in the classified character group according to the number of strokes, ascending order is used for sorting.

6. A data recognition model training system, used to implement any of the methods of claims 1-5, characterized in that: The system comprises: A data acquisition module, used to acquire a fixed training data set, wherein the fixed training data set contains the same font information of all characters; A text comparison module, used to compare all the texts contained in the fixed training data set to obtain a classified text group, wherein the classified text group contains at least one text; The stroke counting module is used to count the number of strokes of each character in the classified character group and sort the characters in the classified character group according to the number of strokes; A text deformation module, used for deforming all the characters in the classified text group to obtain a deformed text data set, wherein the deformed text data set contains a correspondence between each character and the deformed text; The data integration module is used to integrate the deformed text data set and the fixed training data set to obtain an auxiliary training data set, and use the auxiliary training data set to train the data recognition model.

7. The data recognition model training system according to claim 6, characterized in that: The text comparison module includes: A data statistics unit, used for counting the number of strokes of all characters contained in the fixed training data set, and classifying the characters according to the number of strokes to obtain stroke classification character groups, wherein the strokes of the characters in the stroke classification character groups are the same; A feature comparison unit is used to number the stroke classification character groups according to the number of strokes, select a character in a stroke classification character group each time, compare it with other characters in the stroke classification character group, and determine distinguishing stroke features; The data classification unit is used to classify characters whose number of stroke distinguishing features is less than a preset value into one category to obtain a classified character group.

8. The data recognition model training system according to claim 6, characterized in that: The stroke statistics module comprises: A stroke analysis unit, used to count the number of strokes of each character in each classified character group; The stroke sorting unit is used to sort the characters in the classified character group according to the number of strokes.

Images