CN104036290A - Method and device for identifying face value of paper money - Google Patents

Abstract

The invention provides a method and a device for identifying a face value of paper money. The method comprises the steps of: performing projection one-dimensional treatment on an obtained image of a paper money face value characteristic region and performing sliding matching on a template after subjected to the one-dimensional treatment. The device is a face value identification device arranged on an automatic teller machine and mainly comprises a binaryzation module, a projection module and a sliding matching module. The matching algorithm of the invention is to transform two-dimensional image data into a one-dimensional projection curve, and then calculate a correlation coefficient with the template curve. The projection matching based processing way enable the algorithm speed to be greatly improved, and meanwhile, the algorithm is very strong in robustness and very high in recognition rate.

Description

A kind of bank note face amount recognition methods and device

Technical field

The present invention relates to area of pattern recognition, particularly banknote image pattern-recognition, is a kind of method and apparatus of identifying bank note face amount.

Background technology

Paper money recognition comprises that face amount is towards identification and forge or true or paper money discriminating.The identification process that bank note is general is first bank note to be carried out the pre-service such as slant correction, then identify the Currency Type of bank note and face amount towards, then according to the face amount of bank note towards the discriminating of carrying out forge or true or paper money.The face amount of bank note is identified in the identification of bank note occupies very critical role, if first do not identify face amount, is just far from being bank note is carried out to true and false discriminating.

Because bill acceptor system requires to have real-time, General Requirements completes the face amount of bank note and differentiates towards identification and the true and false in 40ms, so just must be optimized processing to recognizer.General bank note face amount recognition methods is the 2-D data processing of bank note data being carried out to full width face, the image of the bank note face amount characteristic area getting is compared from the standard form that is stored in the different bank note face amount characteristic areas in storer, when the image of the bank note face amount characteristic area getting mates correlation and is greater than the threshold value of setting with standard form, be this template face amount, and picture the 5th cover Renminbi direct picture is more similar, process and easily become on the contrary other face amount by mistake like this, so this way discrimination effect is not best, and data processing and recognition time also long, keep for bank note to carry out the false proof time by having shortened like this.

Summary of the invention

The invention provides a kind of recognition methods of bank note face amount and system fast, solve the easily identification by mistake of bank note face amount, the problem that the recognizer time is long.

Technical scheme of the present invention is: the recognition methods of a kind of bank note face amount, the image of the bank note face amount characteristic area getting is compared from the standard form that is stored in the different bank note face amount characteristic areas in storer, when the image of the bank note face amount characteristic area getting mates correlation and is greater than the threshold value of setting with standard form, be this template face amount, comprise the following steps

The binaryzation of steps A, feature regional images;

Step B, the image after binaryzation is done to vertical projection, obtain drop shadow curve;

Step C, select standard form shiding matching in drop shadow curve, pixel value of every slip, adopts the correlation that mates of following formula calculating drop shadow curve and template:

$s_m=\frac{{\mathrm{\Σ}}_{i=0}^{N-1}(f_p\left(i\right)-\stackrel{\‾}{f_p})(f_t\left(i\right)-\stackrel{\‾}{f_t})}{\sqrt{{\mathrm{\Σ}}_{i=0}^{N-1}(f_p\left(i\right)-\stackrel{\‾}{f_p})(f_p\left(i\right)-\stackrel{\‾}{f_p}){\mathrm{\Σ}}_{i=0}^{N-1}(f_t\left(i\right)-\stackrel{\‾}{f_t})(f_t\left(i\right)-\stackrel{\‾}{f_t})}}$

F in formula _p(i) be i point bank note characteristic area binary image drop shadow curve value,

for bank note characteristic area binary image projection average value,

F _t(i) be i point template characteristic area binary image drop shadow curve value,

bank note template characteristic region binaryzation projection average,

S _mit is the coupling correlation while sliding into this pixel;

Step D, ask for s _mmaximal value, be coupling correlation;

Step e, judge whether coupling correlation is greater than decision threshold, if coupling correlation is greater than decision threshold, differentiates for face amount corresponding to bank note template, finishes identifying, otherwise selects next template, repeating step C, D, E.

Further, in above-mentioned bank note face amount recognition methods: described decision gate limit value is 0.7.

Further, in above-mentioned bank note face amount recognition methods: in described step 1, the mode of adoption rate coefficient is carried out binaryzation to feature regional images.

Further, in above-mentioned bank note face amount recognition methods: in the time carrying out vertical projection, the Length Ratio standard form length of drop shadow curve will be grown.

Further, in above-mentioned bank note face amount recognition methods: described feature regional images adopts the false proof area image of ultraviolet reflectance.

The present invention also provides a kind of bank note face amount recognition device, and this device is arranged on automatic teller machine, and the image of the bank note face amount characteristic area getting is identified, and comprises the binarization block of the image of bank note face amount characteristic area being carried out to binaryzation;

Image after binaryzation is done to the projection module of projection;

Standard form;

Standard form in the drop shadow curve that projection module is exported and described standard form carries out the matching module of shiding matching;

The judge module that the coupling correlation of described matching module output is judged.

Further, in above-mentioned bank note face amount recognition device: also comprise the face amount feature regional images cutting module that the bank note face amount feature regional images to obtaining cuts.

Matching algorithm of the present invention is to transfer two-dimensional image data to One Dimensional Projection curve, then calculate the relative coefficient with template curve, this processing mode based on projection matching is greatly improved algorithm speed, and algorithm has very strong robustness simultaneously, and discrimination is very high.

In further mode of the present invention, feature regional images adopts the false proof area image of ultraviolet reflectance, makes the present invention have face amount and identifies and false proof dual-use function.

Below in conjunction with specific embodiment, the present invention is described in more detail.

Brief description of the drawings

Fig. 1 is process flow diagram of the present invention.

Fig. 2 is the vertical projection curve maps of 100 yuans of false proof area images of bill ultraviolet reflectance after binaryzation.

Embodiment

Embodiment 1, the present embodiment is a kind of bank note face amount recognition device being arranged on automatic teller machine, and this device is on automatic teller machine, and the image of the bank note face amount characteristic area that automatic teller machine prior module is got is identified.This bank note face amount recognition device is mainly used to identify bank note characteristic area ultraviolet reflectance gray level image, first, utilize the bank note face amount feature regional images obtaining is cut, obtain face amount recognition image, the face amount recognition image here cuts rear formation by face amount feature regional images cutting module, in binarization block, the image of bank note face amount characteristic area is carried out to binaryzation, the image of the bank note face amount characteristic area of binarization block output carries out binaryzation data and is input to projection module and carries out projection, form drop shadow curve's image of one dimension, mate with the template in standard form successively at matching module, if there is certain coupling correlation to be greater than 0.7, can be identified as corresponding face amount.

In the present embodiment, the bank note face amount recognition methods adopting, mainly comprises the binaryzation of bank note feature regional images, the vertical projection curve of bank note feature regional images and the step such as the sleiding form of drop shadow curve mates.

Before carrying out the identification of bank note face amount, need to cutting bank note feature regional images.The present embodiment is that the false proof area part with fluorescence reaction of ultraviolet reflectance image is cut.

Bank note is entering into after automatic teller machine, in the present embodiment, be mainly to check for 100 yuan of Renminbi, 50 yuan, image acquiring device CIS by automatic teller machine can obtaining 100 yuan, 50 yuans clear picture, taking 100 yuan as example, now, identify bank note towards with towards, according to towards with towards the physical location of the highlighted fluorescence character pattern in recognition result, " 100 " on bank note, feature regional images is cut.Now image may have inclination, therefore, has also carried out slant correction, and then, the former figure of cutting has powerful connections, inclination, after cutting, image is upright.The present embodiment medium dip is proofreaied and correct and is carried out in Cutting feature area image simultaneously.Be better than like this advanced line tilt correction of whole sub-picture, then Cutting feature area image, has saved the regular hour.

Start to carry out after paper money recognition the first binaryzation to bank note feature regional images.The method of image binaryzation has a lot, the image-region of getting for the present invention, " 100 " first image place image of having powerful connections, and the discreteness of banknote image, the inconsistency of new and old banknote image, adopt the bimodal method of histogram, the method texts such as process of iteration are bad, investigate " 100 " and locate image, " 100 " pattern is brighter than background image, and " 100 " pattern magnitude is substantially constant, so adopting proportional Y-factor method Y, the present invention can well distinguish " 100 " pattern and background, scale-up factor method calculated amount is little, binaryzation effect stability, new and old bank note is had to good adaptability.The number percent of scale-up factor is to obtain by artificial experiment, and the choosing of number percent will make image background and picture material after binaryzation reach best.

Next just image after binaryzation is carried out to vertical projection.The length of drop shadow curve is longer than standard form length, and this is the adaptive faculty in order to improve feature regional images paper currency printing position discreteness.As shown in Figure 2, be the vertical projection curve maps of 100 yuans of false proof area images of bill ultraviolet reflectance after binaryzation.

Then be with standard form shiding matching in drop shadow curve, get the maximal value of sleiding form coupling as the net result of coupling.First take out 100 standard form, adopt shiding matching to use following formula to mate correlation value calculation.

Drop shadow curve's sleiding form matching algorithm of the present embodiment, its thought is by the anti-counterfeiting characteristic zone map of the ultraviolet reflectance of well cutting, after image binaryzation, then asks for its horizontal projection or vertical projection, obtains sequence f _p(i), then by the projection sequence in the projection sequence of bank note and the template base made such as the projection of the characteristic area of 100 yuans compares, ask for the correlation of sequence.If when coupling correlation is greater than decision threshold, differentiate for face amount corresponding to bank note template.Projection matching formula of correlation coefficient is:

$s_m=\frac{{\mathrm{\Σ}}_{i=0}^{N-1}(f_p\left(i\right)-\stackrel{\‾}{f_p})(f_t\left(i\right)-\stackrel{\‾}{f_t})}{\sqrt{{\mathrm{\Σ}}_{i=0}^{N-1}(f_p\left(i\right)-\stackrel{\‾}{f_p})(f_p\left(i\right)-\stackrel{\‾}{f_p}){\mathrm{\Σ}}_{i=0}^{N-1}(f_t\left(i\right)-\stackrel{\‾}{f_t})(f_t\left(i\right)-\stackrel{\‾}{f_t})}}$

F in formula _p(i) be i point bank note characteristic area binary image drop shadow curve value,

for bank note characteristic area binary image projection average value,

F _t(i) be i point template characteristic area binary image drop shadow curve value,

bank note template characteristic region binaryzation projection average,

S _mit is the coupling correlation while sliding into this pixel;

In order to improve the adaptive faculty of feature regional images printing position error, adopt shiding matching strategy at this, with make to identify face amount standard form drop shadow curve in bank note drop shadow curve at shiding matching, pixel value of every slip, calculate one time related coefficient S, obtain like this related coefficient sequence s _m, the selection of slip number of times is relevant with the error of paper currency printing position, finally asks for related coefficient sequence s _mmaximal value, as the final matching results of matching algorithm, when the peak value of coupling related coefficient is while being greater than 0.7, can be identified as corresponding face amount.

The present invention extracts the characteristic area of the false proof area image of ultraviolet reflectance as identification bank note face amount, and this greatly reduces the data volume of processing, both bank note has been carried out to face amount identification, has carried out false proof again to bank note.The matching process of bank note face amount identification is to adopt drop shadow curve's coupling, like this image 2-D data is converted to One Dimensional Projection data, finally make algorithm process speed of the present invention be exceedingly fast, the program code of the method is transplanted under TMS320C6416 DSP platform, after tested, identify 2ms consuming time, meet high speed Real-time System, reserved the more time to guiding against false of paper currency.

Claims (7)

1. bank note face amount recognition methods, the image of the bank note face amount characteristic area getting is compared from the standard form that is stored in the different bank note face amount characteristic areas in storer, when the image of the bank note face amount characteristic area getting mates correlation and is greater than the threshold value of setting with standard form, be this template face amount, it is characterized in that: comprise the following steps

The binaryzation of steps A, feature regional images;

Step B, the image after binaryzation is done to vertical projection, obtain drop shadow curve;

Step C, select standard form shiding matching in drop shadow curve, pixel value of every slip, adopts the correlation that mates of following formula calculating drop shadow curve and template:

$s_m=\frac{{\mathrm{\Σ}}_{i=0}^{N-1}(f_p\left(i\right)-\stackrel{\‾}{f_p})(f_t\left(i\right)-\stackrel{\‾}{f_t})}{\sqrt{{\mathrm{\Σ}}_{i=0}^{N-1}(f_p\left(i\right)-\stackrel{\‾}{f_p})(f_p\left(i\right)-\stackrel{\‾}{f_p}){\mathrm{\Σ}}_{i=0}^{N-1}(f_t\left(i\right)-\stackrel{\‾}{f_t})(f_t\left(i\right)-\stackrel{\‾}{f_t})}}$

F in formula _p(i) be i point bank note characteristic area binary image drop shadow curve value,

for bank note characteristic area binary image projection average value,

F _t(i) be i point template characteristic area binary image drop shadow curve value,

bank note template characteristic region binaryzation projection average,

S _mit is the coupling correlation while sliding into this pixel;

Step D, ask for s _mmaximal value, be coupling correlation;

Step e, judge whether coupling correlation is greater than decision threshold, if coupling correlation is greater than decision threshold, differentiates for face amount corresponding to bank note template, finishes identifying, otherwise selects next standard form, repeating step C, D, E.

2. bank note face amount according to claim 1 recognition methods, is characterized in that: described decision gate limit value is 0.7.

3. bank note face amount according to claim 2 recognition methods, is characterized in that: in described step 1, the mode of adoption rate coefficient is carried out binaryzation to feature regional images.

4. bank note face amount according to claim 2 recognition methods, is characterized in that: in the time carrying out vertical projection, the Length Ratio standard form length of drop shadow curve will be grown.

5. according to arbitrary described bank note face amount recognition methods in claim 1 to 4, it is characterized in that: described feature regional images adopts the false proof area image of ultraviolet reflectance.

6. a bank note face amount recognition device, is arranged on automatic teller machine, and the image of the bank note face amount characteristic area getting is identified, and it is characterized in that:

Comprise the binarization block of the image of bank note face amount characteristic area being carried out to binaryzation;

Image after binaryzation is done to the projection module of projection;

Standard form;

Standard form in the drop shadow curve that projection module is exported and described standard form carries out the matching module of shiding matching;

The judge module that the coupling correlation of described matching module output is judged.

7. bank note face amount recognition device according to claim 6, is characterized in that: also comprise the face amount feature regional images cutting module that the bank note face amount feature regional images to obtaining cuts.