CN104463122A - Seal recognition method based on PCNN - Google Patents

Abstract

The invention discloses a seal recognition method based on a PCNN. A corresponding experimental environment is created, and a dialog box program is written based on an MFC framework. The size and format of an image to be processed are regulated, and the image to be processed is eroded and dilated; skeleton extraction is carried out on the seal inscription image based on a model of the PCNN to obtain a corresponding seal inscription skeleton; image fusion is carried out on the obtained seal inscription skeleton and the seal inscription image obtained after binarization to obtain a fused image; the Hu moment value of the seal inscription image to be recognized and the Hu moment value of a seal inscription image obtained by stamping through an office-copy standard seal of a system are calculated, wherein a function according to which the Hu moment values of the images can be calculated exists in an OpenCV library; the similarity of the Hu moments of the two images is calculated, seven normalized central moments are averaged to obtain a relatively accurate threshold value according to which whether the two images are matched can be judged, and whether the seal inscription image is true is recognized according to the threshold value.

Description

The seal recognition methods of a kind of Based PC NN

Technical field

The invention belongs to the applied technical field of Digital Image Processing, relate to the seal recognition methods of a kind of Based PC NN.

Background technology

The retrieval of image and relevant processing and identification, present today of explosive increase in quantity of information, more and more highlight its important effect in the social production and daily life of people.The mankind are mainly directly obtained and exchange information by this medium of image.Along with developing rapidly of computer technology and artificial intelligence, the process of digital picture is just being become to the focus of society and academia's research.Retrieval process mode traditionally for image is the shape of being observed image by human eye, size, the information such as color, the central nervous system then through the mankind makes differentiation to it, discrimination is very high, but it is but difficult to the requirement of satisfied current every application in processing speed.

Carry out processing in artificial intelligence for digital picture, machine learning, Digital Image Processing, the association areas such as computer vision have very important application, become the concern direction of Chinese scholars gradually.In recent years, the model having one to be referred to as the artificial neural network of Pulse Coupled Neural Network PCNN (Pulse Coupled Neural Networks) being applied in image procossing just gradually.PCNN model is a kind of biological network model of closely human brain neural network, and at image procossing, image recognition, decision optimization aspect also exists advantage that is different and traditional artificial neural network, has broad application prospects.It simulates biological visual characteristic, has biological background context, is referred to as third generation neural network.As far back as 1999, the Izhikevich mathematically actual biological cell model of Strict Proof was consistent with PCNN model, the coordinate of different just variable.Research shows, the fundamental characteristics that PCNN has has change threshold property, non-linear modulation characteristic, and synchronizing pulse provides phenomenon, capture characteristic, dynamic pulse provides phenomenon, auto-wave characteristic and comprehensive space-time characterisation, image denoising can be applied to, Iamge Segmentation, Image Edge-Detection etc.The important attribute of the synchronous and auto-wave that PCNN has, obtain the concern of researchist, the relevant treatment for digital picture has good effect.

Seal image is because the attribute of its carrying, so it is generally the place appearing at complex background medium, this adds very large difficulty also to the identification of printed text.Further comprises in its background some other with the structure of printed text image similarity, color and the printed text image of these contents are closely similar, cause the difficulty of printed text identification equally.

Traditionally, for the feature of printed text image, because most printed text image covers out corresponding printed text image by the ink paste of redness, traditional extraction to printed text image, directly the red component in the printed text image of original RGB color space can be extracted from the mixed and disorderly shading of seal and complicated background, then entered the process of binaryzation, in the process of process, remove seal as much as possible background, word etc. affect the noise of image information, keep the more intrinsic information of seal.

Utilize traditional step extracted printed text image as follows:

Step 1: based on the experimental situation of putting up, utilizes the architectural feature of printed text image, extracts the red component of printed text image.

Step 2: first carry out gray proces for original printed text image, then carry out the operation of binaryzation, reduces the redundant information of image, convenient process.

Remaining step operates with the experimental procedure 6,7 of the printed text image procossing of above-mentioned Based PC NN.

Summary of the invention

The object of the invention is to the defect overcoming the existence of above-mentioned technology, the seal recognition methods of a kind of Based PC NN is provided, is based upon in Windows operating system platform.Therefore the printed text treatment research of Based PC NN model can utilize the visual integrating and developing platform Visual C++6.0 of Microsoft.MFC basic framework can be built in this Integrated Development Environment, uses MFC framework to build experimental arrangement, and is aided with this computer vision storehouse of increasing income of OpenCV1.0, by relevant algorithm application to in the process of image.Its concrete technical scheme is:

A seal recognition methods of Based PC NN, comprises the following steps:

Step 1: after putting up corresponding experimental situation, just can based on MFC framework write one we test the dialog box program of needs, be convenient to show experimental result.

Step 2: the operation original printed text image being carried out to binaryzation, reduces the redundant information of image, convenient process.

Step 3: size and stylistic regular is carried out to pending image, and in the operation of corroding it and expand.

Step 4: the model of Based PC NN carries out the extraction operation of skeleton to the image of printed text, obtain corresponding printed text skeleton.

Step 5: the printed text image that the binary conversion treatment that the printed text skeleton obtained according to step 4 and step 2 obtain is crossed, by the process that both carry out image co-registration, obtains fused images.

Step 6: the Hu square value calculating the regular seal lid printed text image out that printed text image to be identified and system are kept on file, having in OpenCV storehouse can the function of value of computed image Hu square.

Step 7: the similarity calculating two width image Hu squares, and 7 normalized center squares are averaging obtain one and relatively accurate can judge the threshold value whether two width images mate, differentiate that printed text image is true and false according to threshold value.

Compared with prior art, beneficial effect of the present invention is:

Computer technology is adopted to mate automatically for printed text image, compare with traditional artificial cognition printed text image, eliminate background media when seal is added a cover, a series of people such as uneven and picture noise that may exist of size of exerting oneself when adding a cover with the naked eye carries out the impact observing and judge to cause.Adopt artificial nerve network model to have high fault tolerance and powerful adaptive faculty that biological neural network has, when carrying out image procossing, Pulse-coupled Neural Network Model is the same with Neural Networks System, according to parallel work-flow, image is processed, this makes it possible to the speed greatly improving image procossing.Simultaneously due to the computing power that computing machine is outstanding, the vision mistaken ideas that the vision overcoming people in artificial cognition may exist, the very close target that such as some human eyes cannot judge.The application of neural network model makes printed text image comparatively complicated at background media, when target image is very similar with the image of background, can be separated accurately and the loss of image information can maximizedly be reduced simultaneously, identification for seal image is more accurate, efficiency can be higher, effectively overcome many difficult problems of traditional seal identification, obtain good result.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the seal recognition methods that the present invention is based on PCNN;

Fig. 2 is matching result figure, and wherein, Fig. 2 a. original image, Fig. 2 b. extracts red component image, Fig. 2 c. fused images.

Embodiment

The technological means realized to make the present invention, creation characteristic, reaching object and effect is easy to understand, below in conjunction with accompanying drawing and instantiation, setting forth the present invention further.

In FIG, what identification compared by computing machine is two width images, and a width is the printed text image of keeping on file being stored in computing machine this locality by technological means, and another width is then the printed text image to be identified needing to compare.

Utilize the program of computing machine to process automatically for two width images, first input the printed text image that two width need to compare.

Conveniently two width printed text images are compared, needed to carry out pre-service to printed text image before printed text image is compared, reduce the information that image is unnecessary, pretreated printed text image the background, word etc. of removal seal as much as possible can affect the noise of image information while retaining original image information to greatest extent, improves the precision of printed text identification.

Be input in PCNN neural network model by printed text image good for pre-service, obtain the printed text skeleton after process refinement, the fusion then carrying out image with the printed text image of binaryzation obtains finally wanting and printed text of keeping on file carries out the fused images of mating.In traditional matching and recognition method, in view of general seal is all using the color of redness as ink paste, therefore in traditional method, usually be all by extracting the simple profile obtaining image to the component of image red channel, and then printed text image is compared, draw matching result, as shown in Figure 2.

Can see in the image obtained based on traditional extraction printed text red component passage and the fused images taking neural network model process to obtain herein.Utilize the function inside the experimental situation of putting up to the Hu of digital picture not bending moment calculate, on the basis of great many of experiments, setting threshold value, draws the similarity of identical printed text image Hu square, finally draws matching result.

After obtaining fused images, the theory of digital picture not bending moment is utilized to carry out match cognization for printed text image.

The main principle of bending moment comes from several squares of the insensitive fundamental region of image conversion as shape facility.Not bending moment not what definition in the physical significance of image, definition just in pure mathematics, when a width consecutive image, supposes that the function of image is f (x, y), so the p+q rank geometric moment (i.e. standard square) of image just can be defined as:

$m_{\mathrm{pq}}={\∫}_{-\∞}^{\∞}{\∫}_{-\∞}^{\∞}{x}^p{y}^{p}f(x,y)\mathrm{dxdy},p,q=\mathrm{0,1,2}....---\left(1\right)$

The centre distance of p+q can be defined as

${\mathrm{\μ}}_{\mathrm{pq}}={\∫}_{-\∞}^{\∞}{\∫}_{-\∞}^{\∞}{(x-\stackrel{\‾}{x})}^p{(y-\stackrel{\‾}{y})}^{q}f(x,y)\mathrm{dxdy},p,q=\mathrm{0,1,2}....---\left(2\right)$

Wherein with the center of gravity of y representative image.

$\stackrel{\‾}{x}=\frac{m_{10}}{m_{00}}---\left(3\right)$

$\stackrel{\‾}{y}=\frac{m_{10}}{m_{00}}---\left(4\right)$

For the digital picture of discrete type, summation number can be adopted to replace integration:

$m_{\mathrm{pq}}={\mathrm{\Σ}}_{y=1}^N{\mathrm{\Σ}}_{x=1}^M{x}^p{y}^{q}f(x,y),p,q=\mathrm{0,1,2}....---\left(5\right)$

${\mathrm{\μ}}_{\mathrm{pq}}={\mathrm{\Σ}}_{y=1}^N{\mathrm{\Σ}}_{x=1}^M{(x-\stackrel{\‾}{x})}^p{(y-\stackrel{\‾}{y})}^{q}f(x,y)\mathrm{dxdy},p,q=\mathrm{0,1,2}....---\left(6\right)$

Here we are height and the width of image to their definition to N and M;

Center square after normalization can be defined as:

${\mathrm{\η}}_{\mathrm{pq}}={\mathrm{\μ}}_{\mathrm{pq}}/\left({\mathrm{\μ}}_{00}^{\mathrm{\ρ}}\right)---\left(7\right)$

Wherein $\mathrm{\ρ}=\frac{(p+q)}{2}+1---\left(8\right)$

Second order and three normalization center, rank squares are utilized to construct 7 not bending moment M1-M7:

M1＝η ₂₀+η ₀₂(9)

M2＝(η ₂₀-η ₀₂) ²+4η11 ²(10)

M3＝(η ₃₀-3η ₁₂) ²+(3η ₂₁-η ₀₃) ²(11)

M4＝(η ₃₀+η ₁₂) ²+(η ₂₁+η ₀₃) ²(12)

M5＝(η ₃₀-3η ₁₂)(η ₃₀+η ₁₂)((η ₃₀+η ₁₂) ²-3(η ₂₁+η ₀₃) ²)

+(3η ₂₁-η ₀₃)(η ₂₁+η ₀₃)(3(η ₃₀+η ₁₂) ²-3(η ₂₁+η ₀₃) ²)

(13)

M6＝(η ₂₀-η ₀₂)(η ₃₀+η ₁₂) ²-(η ₂₁-η ₀₃) ²)+4η ₁₁(η ₃₀+η ₁₂)(η ₂₁+η ₀₃)

(14)

M7＝(3η ₂₁-η ₀₃)(η ₃₀+η ₁₂)((η ₃₀+η ₁₂) ²-3(η ₂₁+η ₀₃) ²)-(η ₃₁-3η ₁₂)(η ₃₀

+η ₁₂)(3(η ₃₀+η ₁₂) ²-(η ₂₁+η ₀₃))

(15)

These 7 not bending moment form a stack features amount, Hu.M.K demonstrated them in 1962 and has rotation, zooming and panning unchangeability.Because every width picture has the square of oneself, so for the coupling of two width images, only need several squares of comparison two width image whether to mate and just can reach the whether identical object of discriminating image.In the application of reality, it is relatively good that the unchangeability of M1 and M2 keeps, and other several not bending moments then have the larger error of existence.By Hu not bending moment image is identified, the velocity ratio that advantage is identifying is very fast, matching process is uncomplicated, but discrimination is lower, therefore for Hu not bending moment use we be generally all in the enterprising row operation of low-order moment, can by for same regular seal print off the statistical computation of difference of Hu square between the printed text image that comes and masterplate printed text image, can obtain when this difference is less than certain threshold value time, whether just can study and judge two width images is come from same piece of seal lid identical image out, just can identify the true and false of printed text image to be identified whereby.

The present invention compared with artificial cognition seal image, Computerized intelligent carry out mating that better can to reduce the efficiency that artificial cognition occurs low to printed text image, the situation that discrimination is low.With general based on only extracting compared with red channel component in colored printed text images match process, remain the related keyword information of printed text image to a greater extent, the accuracy of the coupling of raising.Application neural network model carries out image procossing, has the parallel characteristics of neural network model, greatly improves the speed of Digital Image Processing.For the Hu square of digital picture, it has rotation, zooming and panning unchangeability, and in the matching process of digital picture, make matching process can not too complex, recognition speed be very fast.Choose Pulse-coupled Neural Network Model can adjust neural network model setting parameter according to different situations simultaneously, optimize the process for digital picture, improve speed and the precision of Digital Image Processing.

The above, be only best mode for carrying out the invention, is anyly familiar with those skilled in the art in the technical scope that the present invention discloses, and the simple change of the technical scheme that can obtain apparently or equivalence are replaced and all fallen within the scope of protection of the present invention.

Claims (1)

1. a seal recognition methods of Based PC NN, is characterized in that, comprise the following steps:

Step 1: after putting up corresponding experimental situation, just can with write based on MFC framework one we test the dialog box program of needs, be convenient to show experimental result;

Step 2: the operation original printed text image being carried out to binaryzation, reduces the redundant information of image, convenient process;

Step 3: size and stylistic regular is carried out to pending image, and in the operation of corroding it and expand;

Step 4: the model of Based PC NN carries out the extraction operation of skeleton to the image of printed text, obtain corresponding printed text skeleton;

Step 5: the printed text image that the binary conversion treatment that the printed text skeleton obtained according to step 4 and step 2 obtain is crossed, by the process that both carry out image co-registration, obtains fused images;

Step 6: the Hu square value calculating the regular seal lid printed text image out that printed text image to be identified and system are kept on file, having in OpenCV storehouse can with the function of the value of computed image Hu square;

Step 7: the similarity calculating two width image Hu squares, and 7 normalized center squares are averaging obtain one and relatively accurate to judge the threshold value whether two width images mate, can differentiate that printed text image is true and false according to threshold value.