CN105701488A - Identity card identification method - Google Patents

Abstract

The present invention relates to the image processing technology field, and provides an identity card identification method. The identity card identification method comprises the following steps of acquiring an identity card image, carrying out the pre-processing on the acquired identity card image, carrying out the positioning segmentation on the image after the pre-processing, and identifying the image after the positioning segmentation. According to the present invention, an algorithm adopted in the threshold value binarization processing is simple and practical, so that the identity card identification efficiency and accuracy are high.

Description

A kind of identity card recognition method

Technical field

The present invention relates to technical field of image processing, particularly relate to a kind of identity card recognition method。

Background technology

Identity card is the unique perfect instrument proving citizen's legal identity, it is that citizen participates in certificate necessary to various politics, economy, social activity, such as: entrance activity meeting-place, handle bank card, move in hotel, airplane train, Internet bar's online, handle all kinds of credentialss etc.。It can be said that identity card has goed deep into the various aspects of our life。Therefore, how fast and accurately reading identity card information becomes more and more important。

Along with the application popularization of smart mobile phone, the identity card identification of mobile phone terminal is same with PC end also to grow up therewith。Photographic head also becomes wisdom, recognizable various Word messages。The identity card recognition method of mobile phone, solves the problem that information input is difficult substantially。But owing to mobile phone terminal and PC end aspect of performance vary, recognition efficiency is restricted, recognition accuracy is generally very low。

Summary of the invention

It is an object of the invention to provide a kind of identity card recognition method, it is intended to solve the general very low problem of existing identity card recognition method accuracy rate。

The invention provides a kind of identity card recognition method, comprise the following steps:

Captured identity card image；

The ID Card Image of collection is carried out pretreatment；

Pretreated image is positioned segmentation；

Image after locating segmentation is identified。

Further, the ID Card Image of collection is carried out pretreatment, specifically includes following steps:

Convert the ID Card Image of collection to gray level image；

Gray level image is carried out smoothing denoising process；

Gray level image after being processed by smoothing denoising carries out threshold values binary conversion treatment；

Image after threshold values binary conversion treatment is carried out slant correction。

Further, the gray level image after smoothing denoising process carries out threshold values binary conversion treatment, specifically includes:

Adopt below equation, gray level image carried out threshold values binary conversion treatment:

$g_s\left(n\right)=\underset{i=0}{\overset{s-1}{\Σ}}{(1-\frac{1}{s})}^{i}g(n-i)---\left(1\right)$

$h\left(n\right)=\frac{g\left(n\right)+g(n-width)}{2}---\left(2\right)$

$g_s\left(n\right)=g_s(n-1)-\frac{g_s(n-1)}{s}+h\left(n\right)---\left(3\right)$

$T\left(n\right)=\left\{\begin{array}{cc}1& g\left(n\right)<\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\\ 0& g\left(n\right)\&GreaterEqual;\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\end{array}\right.---\left(4\right)$

G (n) represents the gray value of n-th in present image, and g (n-width) represents the gray value of (n-width) individual point, g_sN () represents the gray value sum of by a certain percentage n-th above s point, T (n) represents the value after binaryzation, and 0 represents white, 1 expression black, and width is picture traverse。

Further, s takes s=width/8, t and takes t=15。

Further, pretreated image positions segmentation to specifically include the Character segmentation to identity card and split with to the head portrait that identity card is corresponding。

Further, the Character segmentation of identity card is comprised the following steps:

Character pixels is carried out expansion process, obtains multiple text block；

Each text block is gone out according to identity card Feature Selection；

Provincial characteristics according to each text block, finds the connected region meeting ID (identity number) card No. provincial characteristics from connected region as the region of ID (identity number) card No., and is split by this area image。

Further, the Character segmentation of identity card is also included: using ID (identity number) card No. regional location as object of reference, find the corresponding region meeting head portrait provincial characteristics, address provincial characteristics, date of birth provincial characteristics, sex provincial characteristics, name region characteristic of field and regional national feature respectively, and corresponding head portrait region, region, address, date of birth region, sex region, name region and regional national are split。

Further, the image after locating segmentation is identified, specifically includes following steps:

Project each literary composition block divided；

Extract each literary composition block feature data；

It is right the character feature data of each literary composition block feature data and template base to be carried out；

Select similarity the highest as recognition result。

Beneficial effect: identity card recognition method of the present invention, owing to the algorithm adopted in threshold values binary conversion treatment is simple, practical so that the efficiency of identity card identification and accuracy rate are all significantly high。

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of a kind of identity card recognition method that one embodiment of the invention provides；

Fig. 2 is the ID Card Image schematic diagram that one embodiment of the invention provides。

Detailed description of the invention

In order to make the purpose of the present invention, technical scheme and advantage clearly understand, below in conjunction with the drawings and specific embodiments, the present invention is described in further detail。

Fig. 1 is the schematic flow sheet of a kind of identity card recognition method that one embodiment of the invention provides；With reference to Fig. 1, embodiments provide a kind of identity card recognition method, comprise the following steps:

Step S101, captured identity card image。

ID Card Image gathers, this technology is developed for mobile phone application, Automatic Program regulates mobile phone camera and carries out automatically focusing on, balancing the operations such as light filling to reach the scanning shoot effect of optimum, recognizer supports 2 kinds of mode photo origins, is that mobile phone focuses on scanning ID Card Image mode and artificial focusing style of shooting automatically respectively。

Step S102, carries out pretreatment by the ID Card Image of collection。

The ID Card Image pretreatment that will gather, specifically includes following steps。

Step S1021, converts the ID Card Image of collection to gray level image。

Gather the RGB image that the ID Card Image returned is colour, and would generally produce much to affect the noise point of mortality in light, focusing etc., therefore before identification, carry out gradation conversion firstly the need of by image。

Step S1022, carries out smoothing denoising process to gray level image。

Gray level image carries out smoothing denoising process, is the process on the noise point much affecting mortality in gray level image。

Step S1023, the gray level image after being processed by smoothing denoising carries out threshold values binary conversion treatment。

Gray level image after smoothing denoising is processed carries out threshold values binary conversion treatment, specifically includes:

Adopt below equation, gray level image carried out threshold values binary conversion treatment:

$g_s\left(n\right)=\underset{i=0}{\overset{s-1}{\&Sigma;}}{(1-\frac{1}{s})}^{i}g(n-i)---\left(1\right)$

$h\left(n\right)=\frac{g\left(n\right)+g(n-width)}{2}---\left(2\right)$

$g_s\left(n\right)=g_s(n-1)-\frac{g_s(n-1)}{s}+h\left(n\right)---\left(3\right)$

$T\left(n\right)=\left\{\begin{array}{cc}1& g\left(n\right)<\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\\ 0& g\left(n\right)\&GreaterEqual;\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\end{array}\right.---\left(4\right)$

G (n) represents the gray value of n-th in present image, and g (n-width) represents the gray value of (n-width) individual point, g_sN () represents the gray value sum of by a certain percentage n-th above s point, T (n) represents the value after binaryzation, and 0 represents white, 1 expression black, and width is picture traverse。

Identification process adopts the basic thought of threshold binarization method: determine the black or white of a pixel, assess threshold values by some meansigma methodss of other points around this point or in scanning sequency just passable。Compare with threshold values and pixel value。First definition g (n) represents the gray value of n-th。T (n) represents the value after binaryzation, and 0 represents white, and 1 represents black, uses g_sN () represents the sum of s the gray value put before n-th, formula representsJust can simply note T (n) with this s and another variable t should be 0 or 1, and this formula is exactly $T\left(n\right)=\left\{\begin{array}{cc}1& g\left(n\right)<\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\\ 0& g\left(n\right)\&GreaterEqual;\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\end{array}\right..$

Optimize further, when T (n) is defined above, be meansigma methods, that is some points scanned before are the same for the impact weight in other words currently put, that is currently the impact of current point is the same by the pixel of 1 pixel distance and the gray value of the pixel of the distance of s-1 pixel. apparent from understanding intuitively, should be from the current pixel put close to more, the impact of current point is more big, more remote then more little. therefore, the embodiment of the present invention is with more properly more efficient substitution value g_s(n), g_sN () is expressed as follows:

$g_s\left(n\right)=\underset{i=0}{\overset{s-1}{\&Sigma;}}{(1-\frac{1}{s})}^{i}g(n-i),$

Here g_sN () is then the sum of gray value by a certain percentage, it can be seen that from this n more close to the proportion of pixel more high, more remote more low. the color obviously so described holding pixel is more accurate. and g here_s(n) and g_s(n-1) just can recurrence being obtained by addition and multiplication, computational efficiency is that comparison is high。

Optimize further, calculating above depends on scanning sequency in fact, that is the definition of this sequence of g (n) is exactly scanning sequency, it is typically all horizontal sweep, like this, pixel value actually depends on the gray value of the abutment points on horizontal level, but does not account for the pixel of vertical direction to associate。Here also there is an explanation, individual g (n) sequence that before can safeguarding, horizontal sweep produces successively, the gray value of horizontal sweep each row is placed in an array, before certain g (n) is used, it is possible to allow the some gray value g (n-width) that is (n-width) puts of gray value g (n) of n point this row corresponding with previous row take a meansigma methods。Wherein, width is picture traverse。It should be noted that, when (n-width) is be more than or equal to 0, corresponding g (n-width) represents the gray value of (n-width) individual point, if (n-width) less than 0, illustrate current nth point corresponding be the gray value of the first row, then g (n-width) can value be 0。

The formula that gray level image carries out threshold binarization algorithm final is as follows:

$g_s\left(n\right)=\underset{i=0}{\overset{s-1}{\&Sigma;}}{(1-\frac{1}{s})}^{i}g(n-i)---\left(1\right)$

$h\left(n\right)=\frac{g\left(n\right)+g(n-width)}{2}---\left(2\right)$

$g_s\left(n\right)=g_s(n-1)-\frac{g_s(n-1)}{s}+h\left(n\right)---\left(3\right)$

$T\left(n\right)=\left\{\begin{array}{cc}1& g\left(n\right)<\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\\ 0& g\left(n\right)\&GreaterEqual;\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\end{array}\right.---\left(4\right)$

G (n) represents the gray value of n-th in present image, and g (n-width) represents the gray value of (n-width) individual point, g_sN () represents the gray value sum of by a certain percentage n-th above s point, T (n) represents the value after binaryzation, and 0 represents white, 1 expression black, and width is picture traverse。

Based on experience value, value best for above-mentioned s and t is to work as s=width/8, t=15, and wherein width is picture traverse, and gray level image carries out the best results of threshold values binary conversion treatment。

Step S1024, carries out slant correction to the image after threshold values binary conversion treatment。

Because when captured identity demonstrate,proves image, gathering the ID Card Image come and be very just not likely to be, here to the image slant correction after threshold values binary conversion treatment, it is possible to improve the accuracy of identity card identification。

Step S103, positions segmentation to pretreated image。

Pretreated image positions segmentation specifically include the Character segmentation to identity card and split with to the head portrait that identity card is corresponding。

Wherein the concrete steps of the Character segmentation of identity card are included: Character segmentation is split with to the head portrait that identity card is corresponding。Wherein the Character segmentation of identity card is comprised the following steps:

Step S1031, carries out expansion process by character pixels, obtains multiple text block。

Expansion process uses algorithm exactly, is expanded at the edge of image。The edge of target or the blank space of inside are filled out by effect exactly。

The algorithm of expansion process: with the structural element of 3*3, the bianry image that each pixel structural element of scanogram covers with it does AND-operation, if being all 0, this pixel of result images is 0。It is otherwise 1, the result of expansion process: make binary image expand。

Step S1032, goes out each text block according to identity card Feature Selection。

Fig. 2 is the ID Card Image schematic diagram that one embodiment of the invention provides, and with reference to Fig. 2 for identity card, its visual useful information generally includes: citizenship number, address, birth, sex, name, nationality, head portrait。

Identity card useful information area image includes: citizenship number information area image 201, certificate address information area image 202, birth information area image 203, gender information's area image 204, name information area image 205, Folk Information area image 206 and head image information area image 207。

Step S1033, the provincial characteristics according to each text block, from connected region, find the region meeting ID (identity number) card No. connected region most as ID (identity number) card No., and this area image is split。

Provincial characteristics according to each text block, determine the information corresponding to each text block, from connected region, find the connected region that the meets region as each text block of identity card, and this area image is split, namely complete the locating segmentation of each literary composition block。

The Character segmentation of identity card is also included: using ID (identity number) card No. regional location as object of reference, find the corresponding region meeting head portrait provincial characteristics, address provincial characteristics, date of birth provincial characteristics, sex provincial characteristics, name region characteristic of field and regional national feature respectively, and corresponding head portrait region, region, address, date of birth region, sex region, name region and regional national are split。

Specifically include the locating segmentation of context below block:

(1). ID (identity number) card No. area image location and segmentation

Area image feature (such as ID (identity number) card No. has continuous print 18 or 17 numerals) according to ID (identity number) card No., the connected region meeting ID (identity number) card No. feature most is found as ID (identity number) card No. region from connected region series, and this area image is split, namely complete location and the segmentation of ID (identity number) card No. area image。

(2). identity certificate address area image location and segmentation

Using ID (identity number) card No. area image position as object of reference, find and meet the region of address area image feature and the region of position feature as address area image, and be partitioned into address area image。

(3). identity card birth area image location and segmentation

Using ID (identity number) card No. area image position as object of reference, find and meet the region of birth area image feature and the region of position feature as birth area image, and be partitioned into birth area image。

(4). identity card sex area image location and segmentation

Using ID (identity number) card No. area image position as object of reference, find and meet the region of sex area image feature and the region of position feature as sex area image, and be partitioned into sex area image。

(5). identity card name region area image location and segmentation

Using ID (identity number) card No. area image position as object of reference, find and meet the region of name region area image feature and the region of position feature as name region area image, and be partitioned into name region area image。

(6). the framing of identity card regional national and segmentation

Using ID (identity number) card No. area image position as object of reference, find and meet the region of regional national characteristics of image and the region of position feature as regional national image, and be partitioned into regional national image。

(7). identity card head portrait area image location and segmentation

Using ID (identity number) card No. area image position as object of reference, find and meet the region of head portrait area image feature and the region of position feature as head portrait area image, and be partitioned into head portrait area image。

Step S104, is identified the image after locating segmentation。

Identity card character is identified, specifically includes following steps:

Project each literary composition block divided；Extract each literary composition block feature data；It is right the character feature data of each literary composition block feature data and template base to be carried out；Select the brief note characteristic of the highest template base of similarity as recognition result。

The software that this method makes can exist as the mode of assembly, it is provided that api interface is to external application, and including mobile terminal and PC end, recognition result is exported by interface。

Output recognition result adopts mobile terminal or the output of PC end。

One or more technical schemes that the embodiment of the present invention provides, have at least techniques below effect:

1) identity card recognition method of the present invention, owing to the algorithm adopted in threshold values binary conversion treatment is simple, practical so that the efficiency of identity card identification and accuracy rate are all significantly high。

2) identity card recognition method of the present invention, not only to numeral, Chinese Character Recognition, is also identified the alike of identity card, improves the accuracy rate of identity card identification。

The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all any amendment, equivalent replacement and improvement etc. made within the spirit and principles in the present invention, should be included within protection scope of the present invention。

Claims (8)

1. an identity card recognition method, it is characterised in that comprise the following steps:

Captured identity card image；

The ID Card Image of collection is carried out pretreatment；

Pretreated image is positioned segmentation；

Image after locating segmentation is identified。

2. identity card recognition method according to claim 1, it is characterised in that described the ID Card Image of collection is carried out pretreatment, specifically comprises the following steps:

Convert the ID Card Image of collection to gray level image；

Gray level image is carried out smoothing denoising process；

Gray level image after being processed by smoothing denoising carries out threshold values binary conversion treatment；

Image after threshold values binary conversion treatment is carried out slant correction。

3. identity card recognition method according to claim 2, it is characterised in that the gray level image after the process of described smoothing denoising carries out threshold values binary conversion treatment, specifically includes:

Adopt below equation, gray level image carried out threshold values binary conversion treatment:

$g_s\left(n\right)=\underset{i=0}{\overset{s-1}{\&Sigma;}}{(1-\frac{1}{s})}^{i}g(n-i)---\left(1\right)$

$h\left(n\right)=\frac{g\left(n\right)+g(n-width)}{2}---\left(2\right)$

$g_s\left(n\right)=g_s(n-1)-\frac{g_s(n-1)}{s}+h\left(n\right)---\left(3\right)$

$T\left(n\right)=\left\{\begin{array}{cc}1& g\left(n\right)<\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\\ 0& g\left(n\right)\&GreaterEqual;\left(\frac{g_s\left(n\right)}{s}\right)\left(\frac{100-t}{100}\right)\end{array}\right.---\left(4\right)$

G (n) represents the gray value of n-th in present image, and g (n-width) represents the gray value of (n-width) individual point, g_sN () represents the gray value sum of by a certain percentage n-th above s point, T (n) represents the value after binaryzation, and 0 represents white, 1 expression black, and width is picture traverse。

4. identity card recognition method according to claim 3, it is characterised in that described s takes s=width/8, t and takes t=15。

5. identity card recognition method according to claim 4, it is characterised in that described pretreated image is positioned segmentation specifically include the Character segmentation to identity card and split with to the head portrait that identity card is corresponding。

6. identity card recognition method according to claim 5, it is characterised in that the described Character segmentation to identity card comprises the following steps:

Character pixels is carried out expansion process, obtains multiple text block；

Each text block is gone out according to identity card Feature Selection；

Provincial characteristics according to each text block, finds the connected region meeting ID (identity number) card No. provincial characteristics from connected region as the region of ID (identity number) card No., and is split by this area image。

7. identity card recognition method according to claim 6, it is characterized in that, the described Character segmentation to identity card also includes: using ID (identity number) card No. regional location as object of reference, find the corresponding region meeting head portrait provincial characteristics, address provincial characteristics, date of birth provincial characteristics, sex provincial characteristics, name region characteristic of field and regional national feature respectively, and corresponding head portrait region, region, address, date of birth region, sex region, name region and regional national are split。

8. identity card recognition method according to claim 7, it is characterised in that described image after locating segmentation is identified, specifically includes following steps:

Project each literary composition block divided；

Extract each literary composition block feature data；

It is right the character feature data of each literary composition block feature data and template base to be carried out；

Select similarity the highest as recognition result。