CN103093227A - Method and device for extracting features of forms - Google Patents

Abstract

The invention provides a method and a device for extracting features of forms, wherein the method comprises the following steps: pre-processing the forms; and extracting the features of images which is used for identifying the types of forms form the pre-processing forms. The device comprises a pre-processing module which is used for conducting the aforesaid procedure and feature extracting module. Furthermore, the image features which are used for identifying the features of forms comprise a SUMX, a SUMA, a SUMB, a SUMC, a SUMD, and a SUME, wherein the SUMX signifies the number of central axes in the forms, SUMA, SUMB, SUMC and SUMD respectively signify the number of central axes in the A, B, C, and D area, the SUME signifies the number of central axes in a rectangular area E, wherein the A, B, C, and D area are at the midpoint positions of widths and heights of the forms, separate the forms into four areas which are two-line and two-row, the areas of the four areas are identical, the rectangular area E and the forms share the same center, and the width and the height of the rectangular area E are one third of the widths and the heights of the forms. By using the method and the device for extracting features of forms, the features of images which are used for identifying types of the forms can be swiftly and accurately extracted from the forms.

Description

Extract method and the device of table features

Technical field

The present invention relates to the form processing technology field, relate in particular to a kind of method and device that extracts table features.

Background technology

At present, in the time of will looking like to classify by the tabular drawing that the equipment such as scanner sweep system storage, the identification of form types is normally realized by the staff.For this reason, the present inventor has proposed a kind of method of automatic identification form types, and with the storage of classifying, a very important step is exactly to extract the type that table features identifies form from form in the method; And the present invention is exactly on the basis of the above, about choosing which type of feature as the feature of sign form types, how to extract the scheme of such feature.

Summary of the invention

In view of this, the invention provides a kind of method and device that extracts table features.Can extract quickly and accurately the characteristics of image of sign form types from form.

The invention provides a kind of method of extracting table features, comprise the steps:

Step a, form is carried out pre-service;

Step b, extract the characteristics of image of sign form types from pretreated form.

further, the characteristics of image of described sign form types comprises: SUMX, SUMA, SUMB, SUMC, SUMD and SUME, SUMX represents axis number in form, SUMA, SUMB, SUMC and SUMD represent respectively A, B, axis number in four zones of C and D, SUME represents the axis number in the E of rectangular area, A wherein, B, four zones of C and D are the wide and high midpoint at form, the zone that four areas that 2 row 2 that form is divided into are listed as equate, rectangular area E has identical center with form, and widely and high be the wide and high by 1/3rd of form.

Further, described step b comprises:

Step b1, extract horizontal line section and vertical line segment from pretreated form, and described step b1 comprises:

Form is corroded in the horizontal direction with horizontal direction straight line line segment structural element, then expands in vertical direction take the expansion texture element as template, the length value of horizontal direction straight line line segment structural element be form width 3/5ths;

Form is corroded in vertical direction with vertical direction straight line line segment structural element, then once expand in the horizontal direction take described expansion texture element as template, the length value of vertical direction straight line line segment structural element be form the cell height 5/7ths.

Further, described step b also comprises:

The horizontal line section that step b2, combining step b1 extract obtains the form framework with vertical line segment;

Step b3, the form framework that step b2 is obtained carry out negate and thinning processing successively;

Extract the characteristics of image of sign form types step b4, the form framework after step b4 processes.

Further, described step b4 comprises:

Axis number SUMX in form framework after step b41, calculating refinement;

Step b42, computation sheet wide and high, in wide and high midpoint, the equal zone of four areas that form is divided into 2 row 2 row: A, B, C and D, and the number of the axis in calculating A, B, C and four locals of D is respectively: SUMA, SUMB, SUMC and SUMD;

Step b43, choose a rectangular area E in form inside, this rectangular area E and form have identical center, and height and width be form height and width 1/3rd, and calculate the number SUME of axis in the E of this rectangular area.

Correspondingly, the present invention also provides a kind of device that extracts table features, comprising:

Pretreatment module is used for form is carried out pre-service;

Characteristic extracting module is used for extracting from pretreated form the characteristics of image that identifies form types.

further, the characteristics of image of the sign form types that described characteristic extracting module is extracted comprises: SUMX, SUMA, SUMB, SUMC, SUMD and SUME, SUMX represents axis number in form, SUMA, SUMB, SUMC and SUMD represent respectively A, B, axis number in four zones of C and D, SUME represents the axis number in the E of rectangular area, A wherein, B, four zones of C and D are the wide and high midpoint at form, the zone that four areas that 2 row 2 that form is divided into are listed as equate, rectangular area E has identical center with form, and widely and high be the wide and high by 1/3rd of form.

Further, described characteristic extracting module comprises:

The line segments extraction unit is used for extracting horizontal line section and vertical line segment from pretreated form, and described line segments extraction unit specifically is used for:

Form is corroded in the horizontal direction with horizontal direction straight line line segment structural element, then expands in vertical direction take the expansion texture element as template, the length value of horizontal direction straight line line segment structural element be form width 3/5ths;

Form is corroded in vertical direction with vertical direction straight line line segment structural element, then once expand in the horizontal direction take described expansion texture element as template, the length value of vertical direction straight line line segment structural element be form the cell height 5/7ths.

Further, described characteristic extracting module also comprises:

The line segment merge cells is used for merging the horizontal line section and vertical line segment that the line segments extraction module is extracted, and obtains the form framework;

Negate and thinning processing unit, the form framework that is used for the line segment merge cells is obtained carries out negate and thinning processing successively;

Feature extraction unit is extracted the characteristics of image that identifies form types for the form framework after negate and thinning processing cell processing.

Further, described feature extraction unit specifically is used for:

Axis number SUMX in form framework after the calculating refinement;

Computation sheet wide and high in wide and high midpoint, is divided into the zone that four areas of 2 row 2 row equate to form: A, B, C and D, and the number of calculating the axis in A, B, C and four locals of D is respectively: SUMA, SUMB, SUMC and SUMD;

Choose a rectangular area E in form inside, this rectangular area E and form have identical center, and height and width be form height and width 1/3rd, and calculate the number SUME of axis in the E of this rectangular area.

Beneficial effect of the present invention:

By form is carried out pre-service, then extract sign form class characteristics of image wherein, can extract quickly and accurately the characteristics of image of sign form types from form, to be conducive to further according to this characteristics of image the form storage of classifying.

Further, select SUMX, SUMA, SUMB, SUMC, SUMD and SUME as the characteristics of image of sign form types.These features can be reacted well the design feature of dissimilar form and are easy to extract from form, therefore when according to these features, form being classified, can guarantee the accuracy of classifying.

further, when the characteristics of image of the characteristics of image that extracts the sign form types, wherein a step is the horizontal line section and vertical line segment that extracts in form, and when demonstrate,proving through repeatedly testing, when extracting horizontal line section, the length value that adopts horizontal direction straight line line segment structural element be form width 3/5ths, when extracting vertical line segment, the length value that adopts vertical direction straight line line segment structural element is 5/7ths o'clock of cell height of form, can guarantee to extract well horizontal line section and vertical line segment, thereby guarantee the SUMX of extraction, SUMA, SUMB, SUMC, SUMD and SUME accuracy.

Description of drawings

The invention will be further described below in conjunction with drawings and Examples:

Fig. 1 is the schematic flow sheet of embodiment of the method for extraction table features of the present invention.

Fig. 2 is the schematic flow sheet of the embodiment of step S11 in Fig. 1.

Fig. 3 is the structural representation of form.

Fig. 4 is the pretreated structural representation of Fig. 3.

Fig. 5 is the horizontal line section that extracts from Fig. 3.

Fig. 6 is the vertical line segment that extracts from Fig. 3.

Fig. 7 is the structural representation of the form that obtains after Fig. 5 and Fig. 6 merge.

Fig. 8 is the structural representation after Fig. 7 negate.

Fig. 9 is the structural representation of the axis that obtains after Fig. 8 refinement.

Figure 10 is the structural representation of embodiment of the device of extraction table features of the present invention.

Figure 11 is the structural representation of the embodiment of the characteristic extracting module in Figure 10.

Embodiment

Please refer to Fig. 1, is the schematic flow sheet of embodiment of the method for extraction table features provided by the invention.The method comprises:

Step S11, form is carried out pre-service.

Wherein, pre-service includes but not limited to: cut apart, binaryzation and filtering processes.Particularly, at first form is carried out dividing processing extraction pure tabular drawing picture wherein, namely remove form word segment on every side.Then pure tabular drawing is looked like to carry out binary conversion treatment, obtain binary image; Preferably, adopt the local binarization method that pure tabular drawing is looked like to process, the step of local binarization method mainly comprises: the threshold value of the first, calculating every bit: $M=\underset{\underset{-d<l<d}{-d<k<d}}{\max }g(x+k,y+l),$ $N=\underset{\underset{-d<l<d}{-d<k<d}}{\min }g(x+k,y+l),$ $T(x,y)=\left\{\begin{array}{c}(M+N)/2,M-N\&GreaterEqual;S\\ T^{\&prime;},M-N<S\end{array}\right.,$ The second, pointwise binaryzation: $b(x,y)=\left\{\begin{array}{c}0,g(x,y)\&le;T(x,y)\\ 1,g(x,y)>T(x,y)\end{array}\right..$ Wherein, g (x, y) denotation coordination (x, y) gray-scale value of locating, the result of b (x, y) expression g (x, y) binaryzation, T (x, y) expression binary-state threshold, (2d+1) * (2d+1) for asking for the template window of threshold value, S, T ' they are a certain critical value, span is [0,128].At last, the binary image of gained is carried out filtering process, remove the noise in the tabular drawing picture, obtain the denoising image; Preferably, adopt " spiced salt " noise in median filter method removal tabular drawing picture, certainly also do not get rid of the modes such as maximal value filtering, mini-value filtering and revised Alpha's mean filter of employing and remove noise.

Step S12, extract the characteristics of image of sign form types from the pretreated form of step S11.

Wherein, the characteristics of image of sign form types comprises: SUMX, SUMA, SUMB, SUMC, SUMD and SUME.The below illustrates implication and the step S12 embodiment of SUMX, SUMA, SUMB, SUMC, SUMD and SUME.As shown in Figure 2, step S12 comprises:

Step S21, extract horizontal line section and vertical line segment from pretreated form.

Wherein, the extraction of horizontal line section and vertical line segment mainly comprises with straight line line segment structural element and corrodes and expanded for two steps with the expansion texture element.Through repeatedly testing and verifying, the below introduces a kind of the have horizontal line section of better effects and the extracting mode of vertical line segment: for the extraction of horizontal line section, corrode in the horizontal direction with horizontal direction straight line line segment structural element, then once expand in vertical direction take the expansion texture element as template; Wherein the length value of horizontal direction straight line line segment structural element is a relative value, is not absolute value, and the image of different machines, different batches scanning may be different, thus length value be one than the ratio value of form length; Consider that the character in form might connect together, in order to extract horizontal line section in the process of corrosion, therefore through experiment when to draw the length value of choosing horizontal direction straight line line segment structural element be 3/5ths left and right of form width effect better; The expansion of horizontal line section being done vertical direction is because the situation of line segment fracture may occur when corroding, and get up for the segment link that will rupture this moment, just need to carry out expansion process, namely the line segment overstriking that extracts; Line segment overstriking one circle is got final product, so the selecting structure element is: $\left[\begin{array}{ccc}1& 1& 1\\ 1& 1& 1\\ 1& 1& 1\end{array}\right].$ For the extraction of vertical line segment, corrode in vertical direction with vertical direction straight line line segment structural element, then once expand in the horizontal direction take the expansion texture element as template.On vertical direction, extraction and the horizontal direction of form straight line line segment are similar, and be better when the length value that draws vertical line line segment structural element through experiment is 5/7ths left and right of table cell height; Line segment and horizontal direction on vertical direction are similar, choose the expansion texture element to be: $\left[\begin{array}{ccc}1& 1& 1\\ 1& 1& 1\\ 1& 1& 1\end{array}\right].$

Horizontal line section and vertical line segment that step S22, combining step S21 extract obtain the form framework.

Step S23, the form framework that step S22 is obtained carry out negate and thinning processing successively.

SUMX, SUMA, SUMB, SUMC, SUMD and SUME in step S24, the form framework of extraction after step S23 processes.

Wherein, at first calculate axis number (that is: table cell number) SUMX in form framework after refinement in this step.Then, the wide W of computation sheet and high H, in wide and high midpoint, tabular drawing is looked like to be divided into the equal zone of four areas of 2 row 2 row: A, B, C and D, and the number of the axis in calculating A, B, C and four locals of D is respectively: SUMA, SUMB, SUMC and SUMD.At last, choose a rectangular area E in form inside, this rectangular area E and form have identical center, and height and width be form height and width 1/3rd, and calculate the number SUME of axis in the E of this rectangular area.Obtain thus the characteristics of image F=(SUMX of form, SUMA, SUMB, SUMC, SUMD, SUME).

Fig. 3-9 have illustrated respectively form, pretreated form, the horizontal line section that extracts, the vertical line segment that extracts, horizontal line section and vertical line segment are merged from form from form form framework, the form framework after negate and the structural representation of the form framework after thinning processing can supply those skilled in the art's reference.

The below introduces the embodiment of the device of extraction table features of the present invention.

Please refer to Figure 10, is the structural representation of embodiment of the device of extraction table features provided by the invention.This device comprises:

Pretreatment module 1 is used for form is carried out pre-service.

Wherein, pre-service includes but not limited to: cut apart, binaryzation and filtering processes.Particularly, at first pretreatment module 1 carries out dividing processing extraction pure tabular drawing picture wherein to form, namely removes form word segment on every side.Then pure tabular drawing is looked like to carry out binary conversion treatment, obtain binary image; Preferably, adopt the local binarization method that pure tabular drawing is looked like to process, the mode of local binarization method mainly comprises: the threshold value of the first, calculating every bit: $M=\underset{\underset{-d<l<d}{-d<k<d}}{\max }g(x+k,y+l),$ $N=\underset{\underset{-d<l<d}{-d<k<d}}{\min }g(x+k,y+l),$ $T(x,y)=\left\{\begin{array}{c}(M+N)/2,M-N\&GreaterEqual;S\\ T^{\&prime;},M-N<S\end{array}\right.,$ The second, pointwise binaryzation: $b(x,y)=\left\{\begin{array}{c}0,g(x,y)\&le;T(x,y)\\ 1,g(x,y)>T(x,y)\end{array}\right..$ Wherein, g (x, y) denotation coordination (x, y) gray-scale value of locating, the result of b (x, y) expression g (x, y) binaryzation, T (x, y) expression binary-state threshold, (2d+1) * (2d+1) for asking for the template window of threshold value, S, T ' they are a certain critical value, span is [0,128].At last, the binary image of gained is carried out filtering process, remove the noise in the tabular drawing picture, obtain the denoising image; Preferably, adopt " spiced salt " noise in median filter method removal tabular drawing picture, certainly also do not get rid of the modes such as maximal value filtering, mini-value filtering and revised Alpha's mean filter of employing and remove noise.

Characteristic extracting module 2 is used for extracting from the pretreated form of pretreatment module 1 characteristics of image that identifies form types.

Wherein, the characteristics of image of sign form types comprises: SUMX, SUMA, SUMB, SUMC, SUMD and SUME.The below illustrates the implication of SUMX, SUMA, SUMB, SUMC, SUMD and SUME and the embodiment of characteristic extracting module 2.As shown in figure 11, characteristic extracting module 2 comprises:

Line segments extraction unit 21 is used for extracting horizontal line section and vertical line segment from pretreated form.

Wherein, the extraction of horizontal line section and vertical line segment mainly comprises with straight line line segment structural element and corrodes and expanded for two steps with the expansion texture element.Through repeatedly testing and verifying, the below introduces a kind of the have horizontal line section of better effects and the extracting mode of vertical line segment: for the extraction of horizontal line section, corrode in the horizontal direction with horizontal direction straight line line segment structural element, then once expand in vertical direction take the expansion texture element as template; Wherein the length value of horizontal direction straight line line segment structural element is a relative value, is not absolute value, and the image of different machines, different batches scanning may be different, thus length value be one than the ratio value of form length; Consider that the character in form might connect together, in order to extract horizontal line section in the process of corrosion, therefore through experiment when to draw the length value of choosing horizontal direction straight line line segment structural element be 3/5ths left and right of form width effect better; The expansion of horizontal line section being done vertical direction is because the situation of line segment fracture may occur when corroding, and get up for the segment link that will rupture this moment, just need to carry out expansion process, namely the line segment overstriking that extracts; Line segment overstriking one circle is got final product, so the selecting structure element is: $\left[\begin{array}{ccc}1& 1& 1\\ 1& 1& 1\\ 1& 1& 1\end{array}\right].$ For the extraction of vertical line segment, corrode in vertical direction with vertical direction straight line line segment structural element, then once expand in the horizontal direction take the expansion texture element as template.On vertical direction, extraction and the horizontal direction of form straight line line segment are similar, and be better when the length value that draws vertical line line segment structural element through experiment is 5/7ths left and right of table cell height; Line segment and horizontal direction on vertical direction are similar, choose the expansion texture element to be: $\left[\begin{array}{ccc}1& 1& 1\\ 1& 1& 1\\ 1& 1& 1\end{array}\right].$

Line segment merge cells 22, horizontal line section and vertical line segment for merging line segments extraction unit 21 extractions obtain the form framework.

Negate and thinning processing unit 23, the form framework that is used for line segment merge cells 22 is obtained carries out negate and thinning processing successively.

Feature extraction unit 24 is for SUMX, the SUMA, SUMB, SUMC, SUMD and the SUME that extract the form framework after negate and thinning processing unit 23 processing.

Wherein, at first feature extraction unit 24 calculates axis number (that is: the table cell number) SUMX in form framework after refinement.Then, the wide W of computation sheet and high H, in wide and high midpoint, tabular drawing is looked like to be divided into the equal zone of four areas of 2 row 2 row: A, B, C and D, and the number of the axis in calculating A, B, C and four locals of D is respectively: SUMA, SUMB, SUMC and SUMD.At last, choose a rectangular area E in form inside, this rectangular area E and form have identical center, and height and width be form height and width 1/3rd, and calculate the number SUME of axis in the E of this rectangular area.Obtain thus the characteristics of image F=(SUMX of form, SUMA, SUMB, SUMC, SUMD, SUME).

Explanation is at last, above embodiment is only unrestricted in order to technical scheme of the present invention to be described, although with reference to preferred embodiment, the present invention is had been described in detail, those of ordinary skill in the art is to be understood that, can modify or be equal to replacement technical scheme of the present invention, and not breaking away from aim and the scope of technical solution of the present invention, it all should be encompassed in the middle of claim scope of the present invention.

Claims (10)

1. a method of extracting table features, is characterized in that: comprise the steps:

Step a, form is carried out pre-service;

Step b, extract the characteristics of image of sign form types from pretreated form.

2. the method for extraction table features as claimed in claim 1, it is characterized in that: the characteristics of image of described sign form types comprises: SUMX, SUMA, SUMB, SUMC, SUMD and SUME, SUMX represents axis number in form, SUMA, SUMB, SUMC and SUMD represent respectively A, B, axis number in four zones of C and D, SUME represents the axis number in the E of rectangular area, A wherein, B, four zones of C and D are the wide and high midpoint at form, the zone that four areas that 2 row 2 that form is divided into are listed as equate, rectangular area E has identical center with form, and widely and high be the wide and high by 1/3rd of form.

3. the method for extraction table features as claimed in claim 1 or 2, it is characterized in that: described step b comprises:

Step b1, extract horizontal line section and vertical line segment from pretreated form, and described step b1 comprises:

Form is corroded in the horizontal direction with horizontal direction straight line line segment structural element, then expands in vertical direction take the expansion texture element as template, the length value of horizontal direction straight line line segment structural element be form width 3/5ths;

Form is corroded in vertical direction with vertical direction straight line line segment structural element, then once expand in the horizontal direction take described expansion texture element as template, the length value of vertical direction straight line line segment structural element be form the cell height 5/7ths.

4. the method for extraction table features as claimed in claim 3, it is characterized in that: described step b also comprises:

The horizontal line section that step b2, combining step b1 extract obtains the form framework with vertical line segment;

Step b3, the form framework that step b2 is obtained carry out negate and thinning processing successively;

Extract the characteristics of image of sign form types step b4, the form framework after step b4 processes.

5. the method for extraction table features as claimed in claim 4, it is characterized in that: described step b4 comprises:

Axis number SUMX in form framework after step b41, calculating refinement;

Step b42, computation sheet wide and high, in wide and high midpoint, the equal zone of four areas that form is divided into 2 row 2 row: A, B, C and D, and the number of the axis in calculating A, B, C and four locals of D is respectively: SUMA, SUMB, SUMC and SUMD;

Step b43, choose a rectangular area E in form inside, this rectangular area E and form have identical center, and height and width be form height and width 1/3rd, and calculate the number SUME of axis in the E of this rectangular area.

6. device that extracts table features is characterized in that: comprising:

Pretreatment module is used for form is carried out pre-service;

Characteristic extracting module is used for extracting from pretreated form the characteristics of image that identifies form types.

7. the device of extraction table features as claimed in claim 6, it is characterized in that: the characteristics of image of the sign form types that described characteristic extracting module is extracted comprises: SUMX, SUMA, SUMB, SUMC, SUMD and SUME, SUMX represents axis number in form, SUMA, SUMB, SUMC and SUMD represent respectively A, B, axis number in four zones of C and D, SUME represents the axis number in the E of rectangular area, A wherein, B, four zones of C and D are the wide and high midpoint at form, the zone that four areas that 2 row 2 that form is divided into are listed as equate, rectangular area E has identical center with form, and widely and high be the wide and high by 1/3rd of form.

8. the device of extraction table features as described in claim 6 or 7, it is characterized in that: described characteristic extracting module comprises:

The line segments extraction unit is used for extracting horizontal line section and vertical line segment from pretreated form, and described line segments extraction unit specifically is used for:

Form is corroded in the horizontal direction with horizontal direction straight line line segment structural element, then expands in vertical direction take the expansion texture element as template, the length value of horizontal direction straight line line segment structural element be form width 3/5ths;

Form is corroded in vertical direction with vertical direction straight line line segment structural element, then once expand in the horizontal direction take described expansion texture element as template, the length value of vertical direction straight line line segment structural element be form the cell height 5/7ths.

9. the device of extraction table features as claimed in claim 8, it is characterized in that: described characteristic extracting module also comprises:

The line segment merge cells is used for merging the horizontal line section and vertical line segment that the line segments extraction module is extracted, and obtains the form framework;

Negate and thinning processing unit, the form framework that is used for the line segment merge cells is obtained carries out negate and thinning processing successively;

Feature extraction unit is extracted the characteristics of image that identifies form types for the form framework after negate and thinning processing cell processing.

10. the device of extraction table features as claimed in claim 9, it is characterized in that: described feature extraction unit specifically is used for:

Axis number SUMX in form framework after the calculating refinement;

Computation sheet wide and high in wide and high midpoint, is divided into the zone that four areas of 2 row 2 row equate to form: A, B, C and D, and the number of calculating the axis in A, B, C and four locals of D is respectively: SUMA, SUMB, SUMC and SUMD;

Choose a rectangular area E in form inside, this rectangular area E and form have identical center, and height and width be form height and width 1/3rd, and calculate the number SUME of axis in the E of this rectangular area.

Images