CN104376300A - Identification method used for intelligent matching of incomplete Chinese characters on basis of grid characteristics - Google Patents

Abstract

The invention discloses an identification method used for intelligent matching of incomplete Chinese characters on the basis of grid characteristics. The method comprises the following steps that S1, a shredded paper restored picture is turned into a 0-1 matrix; S2, the picture positions of the Chinese characters are located through the circulation method that sub-matrices having the same sizes as complete characters are used line by line and column by column; S3, grid partitioning is conducted on the incomplete Chinese characters obtained in the step S2 so as to obtain a plurality of sub-matrices, and characteristics are extracted; S4, intelligent matching identification is conducted on each grid sub-matrix characteristic of a partitioned incomplete Chinese character grid through a standard word bank. The identification method used for intelligent matching of the incomplete Chinese characters on the basis of the grid characteristics solves the problem that even though identification and matching are conducted through machines in the shredded paper restoration technology, incomplete Chinese characters cannot be identified finally due to the situation that errors occur in splicing in lines and columns is solved.

Description

A Recognition Method for Intelligently Matching Incomplete Chinese Characters Based on Grid Features

technical field

The invention relates to a recognition method for intelligently matching incomplete Chinese characters based on grid features.

Background technique

Today, shredded paper recovery technology plays an important role in important fields such as judicial evidence recovery, historical document restoration, and military intelligence acquisition. Shredded paper recovery technology should also be taken into consideration when processing private information.

As shown in Figure 1 and Figure 2, the current shredded paper restoration technology mainly uses a splicing algorithm to store Chinese characters in matrix form by pixel, and restore shredded paper according to the matching degree of paper margins and Chinese characters. Although this method is scientific and easy to implement, there are errors in the recognition and matching by the machine, and in the splicing of rows and columns, which will eventually lead to the problem that Chinese characters cannot be recognized.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art, to provide a recognition method for intelligently matching incomplete Chinese characters based on grid features, and to solve the problem that although the shredded paper recovery technology is recognized and matched by a machine, errors in the splicing of rows and columns lead to failure in the end. The problem of recognizing incomplete Chinese characters.

The purpose of the present invention is achieved through the following technical solutions: a recognition method based on grid features intelligently matching incomplete Chinese characters, comprising the following steps:

S1: Convert the shredded paper recovery map into a 0-1 matrix;

S2: According to the image position positioning rules, use the sub-matrix of the complete character size (the size depends on the average character size in the image) to locate the image position of the Chinese character row by row and column by row;

S3: dividing the incomplete Chinese characters obtained in step S2 into sub-matrix by grid block, and extracting features;

S4: Carry out intelligent matching and recognition on the features of each grid sub-matrix after grid segmentation of the incomplete Chinese characters through the standard lexicon.

Step S1 uses MATLAB software to convert the shredded paper restoration map.

The image location positioning rules described in step S2 include:

(1) If there is a sub-matrix with a width/length equal to the size of a word in the sub-matrix of the complete word size, determine an incomplete word and record the position at the same time;

(2) If there is a sub-matrix with a width/length greater than one word size in the complete word size sub-matrix, it will be determined as an incomplete word, and the position will be recorded at the same time, and then cycle from left to right/up and down respectively, and then an incomplete character will be determined word, and record the position at the same time;

(3) If the sub-matrix of the complete word size contains a sub-matrix whose width/length is less than one word size, it is determined as one incomplete word, and the position is recorded at the same time.

Described step S3 comprises the following sub-steps:

S31: divide the incomplete Chinese characters into multiple sub-matrices according to the size of the incomplete Chinese characters;

S32: Analyze each sub-matrix by wavelet function to extract multiple parameter matrices of the multiple sub-matrix pictures, and use the multiple parameter matrices together as the feature of the incomplete character.

A recognition method for intelligently matching incomplete Chinese characters based on grid features also includes a sub-step of establishing a standard lexicon: decomposing each font size of each complete Chinese character into grids to obtain multiple sub-rectangles of standard features and their multiple Parameter matrix to determine the eigenvalues of a complete Chinese character.

The sub-matrix is a sub-matrix with a size of 2*2.

The parameter matrix includes three parameter matrices of vertical attribute, horizontal attribute and diagonal attribute.

The multiple font sizes mentioned are 8 font sizes between the 10th and the 22nd.

Described step S4 comprises the following sub-steps:

S41: Comparing the plurality of grid sub-matrices obtained in step S3 with the standard feature matrix of each complete Chinese character in the standard lexicon;

S42: If the similarity is greater than a certain ratio, determine that the incomplete character is the complete character in the thesaurus.

A certain ratio described in step S42 is 50%.

The beneficial effects of the present invention are: firstly, the present invention converts the shredded paper restoration image into a 0-1 matrix, and then uses the sub-matrix of the complete character size to locate the image position of the Chinese character row by row according to the image position positioning rule. Judging whether it may be an incomplete character, if possible, save it, and then extract the Chinese character feature vector based on the wavelet function to realize the recognition of the Chinese character in the thesaurus. The invention solves the problem that although the shredded paper recovery technology is recognized and matched by a machine, the incomplete Chinese characters cannot be recognized due to errors in the splicing of rows and columns, and provides a method for recognizing incomplete Chinese characters. the

Description of drawings

Figure 1 is a sample diagram of business correspondence;

Figure 2 is the restoration effect diagram of the sample shredded paper;

Fig. 3 is a flow chart of the method of the present invention.

Detailed ways

The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings: as shown in Figure 3, a kind of recognition method based on grid feature intelligently matching incomplete Chinese characters comprises the following steps:

S1: Convert the shredded paper recovery map into a 0-1 matrix;

S2: Use the sub-matrix of the full word size (the size depends on the average word size in the image) to locate the image position of the Chinese character row by row and column by row;

S3: dividing the incomplete Chinese characters obtained in step S2 into sub-matrix by grid block, and extracting features;

S4: Carry out intelligent matching and recognition on the features of each grid sub-matrix after grid segmentation of the incomplete Chinese characters through the standard lexicon.

Step S1 uses MATLAB software to convert the shredded paper restoration map.

The rule of the image position of the positioning Chinese character described in the step S2 comprises the following substeps:

S21: If there is a sub-matrix with a width/length equal to the size of a word in the sub-matrix of the full word size, determine an incomplete word and record the position at the same time;

S22: If the sub-matrix of the full word size has a width/length greater than one word size, determine it as an incomplete character, record the position at the same time, and cycle from left to right/up and down respectively, and then determine an incomplete character , and record the position at the same time;

S23: If the sub-matrix of the complete word size contains a sub-matrix whose width/length is less than one word size, determine it as one incomplete word, and record the position at the same time.

Described step S3 comprises the following sub-steps:

S31: divide the incomplete Chinese characters into multiple sub-matrices according to the size of the incomplete Chinese characters;

S32: Analyze each sub-matrix by wavelet function to extract multiple parameter matrices of the multiple sub-matrix pictures, and use the multiple parameter matrices together as the feature of the incomplete character.

A recognition method for intelligently matching incomplete Chinese characters based on grid features also includes a sub-step of establishing a standard lexicon: decomposing each font size of each complete Chinese character into grids to obtain multiple sub-rectangles of standard features and their multiple Parameter matrix to determine the eigenvalues of a complete Chinese character.

The sub-matrix is a sub-matrix with a size of 2*2.

The parameter matrix includes three parameter matrices of vertical attribute, horizontal attribute and diagonal attribute.

The multiple font sizes mentioned are 8 font sizes between the 10th and the 22nd.

Described step S4 comprises the following sub-steps:

S41: Comparing the plurality of grid sub-matrices obtained in step S3 with the standard feature matrix of each complete Chinese character in the standard lexicon;

S42: If the similarity is greater than a certain ratio, determine that the incomplete character is the complete character in the thesaurus.

A certain ratio described in step S42 is 50%. the

Claims (10)

1., based on a recognition methods for the incomplete Chinese character of grid search-engine Intelligent Matching, it is characterized in that: it comprises the following steps:

S1: shredded paper palinspastic map is changed into 0-1 matrix;

S2: according to picture position locating rule, locates the picture position of Chinese character by the submatrix round-robin method line by line of complete word size;

S3: the incomplete Chinese character obtained in step S2 is carried out grid piecemeal and is divided into submatrix, extracts feature;

S4: respectively by standard dictionary, Intelligent Matching identification is carried out to each the grid submatrix feature after incomplete Chinese character mesh segmentation.

2. a kind of recognition methods based on the incomplete Chinese character of grid search-engine Intelligent Matching according to claim 1, is characterized in that: step S1 adopts MATLAB software to transform shredded paper palinspastic map.

3. a kind of recognition methods based on the incomplete Chinese character of grid search-engine Intelligent Matching according to claim 1, is characterized in that: the picture position locating rule described in step S2 comprises:

(1): if equal a word size containing width/length in the submatrix of complete word size, then determine an incomplete word, record position simultaneously;

(2) if be greater than a word size containing width/length in the submatrix of complete word size, be then defined as 1 incomplete word, simultaneously record position, and more respectively from left and right/upper and lower two circulations in the other direction, then determine an incomplete word, record position simultaneously;

(3) if be less than a word size containing width/length in the submatrix of complete word size, 1 incomplete word is defined as, simultaneously record position.

4. a kind of recognition methods based on the incomplete Chinese character of grid search-engine Intelligent Matching according to claim 1, is characterized in that: described step S3 comprises following sub-step:

S31: according to incomplete Chinese character size, incomplete Chinese character is divided into multiple submatrix;

S32: the multiple parameter matrixs each submatrix being extracted respectively to this multiple submatrix picture with wavelet function analysis, using this multiple parameter matrix together as the feature of this incomplete word.

5. a kind of recognition methods based on the incomplete Chinese character of grid search-engine Intelligent Matching according to claim 1, it is characterized in that: it also comprises a Criterion dictionary sub-step: by the multiple font size of each complete Chinese character, carry out grid decomposition respectively, obtain the multiple sub-rectangle of standard feature, extract multiple parameter matrixs of this multiple submatrix with wavelet function, determine the eigenwert of a complete Chinese character.

6. a kind of recognition methods based on the incomplete Chinese character of grid search-engine Intelligent Matching according to claim 4 or 5, is characterized in that: described submatrix is the submatrix of 2*2 size.

7. a kind of recognition methods based on the incomplete Chinese character of grid search-engine Intelligent Matching according to claim 4 or 5, is characterized in that: described parameter matrix comprises 3 parameter matrixs of vertical attribute, level property and diagonal angle attribute.

8. a kind of recognition methods based on the incomplete Chinese character of grid search-engine Intelligent Matching according to claim 5, is characterized in that: described multiple font size is 8 kinds of font sizes between No. 10 word to 22 words.

9. a kind of recognition methods based on the incomplete Chinese character of grid search-engine Intelligent Matching according to claim 1, is characterized in that: described step S4 comprises following sub-step:

S41: in the multiple grid submatrix obtained by step S3 and standard dictionary, the standard feature matrix of each complete Chinese character compares;

S42: if similarity is greater than a certain ratio, just judges that this incomplete word is as the complete word of this in dictionary.

10. a kind of recognition methods based on the incomplete Chinese character of grid search-engine Intelligent Matching according to claim 9, is characterized in that: a certain ratio described in step S42 is 50 percent.