CN104658097B - A kind of rmb paper currency denomination identifying method of Histogram Matching based on image - Google Patents

Abstract

The invention provides a kind of RMB denomination identifying method of Histogram Matching based on image, applied to rmb paper currency denomination identifying, wherein, training process includes：Obtain the two-dimensional matrix of image training sample；Every image training sample is filtered；Extract the first histogram vectors of the redgreenblue component of every image training sample；The histogrammic average value for obtaining each color component in multiple image training samples respectively is used as the template of this kind of denominations.Identification process is specifically included：Obtain the two-dimensional matrix of images to be recognized；Images to be recognized is filtered；Extract the second histogram of the redgreenblue component with identification image, the distance between the second histogram and each denominations template are obtained respectively and are compared, the minimum corresponding denomination of denominations template of the histogram of distance second is the denomination for being determined as bank note to be identified, and the method that the present invention is provided substantially increases accuracy of identification while ensure that to noise robustness and recognition time is short.

Description

A kind of rmb paper currency denomination identifying method of Histogram Matching based on image

Technical field

Know the present invention relates to the rmb paper currency denomination of financial field, more particularly to a kind of Histogram Matching based on image Other banknote denomination recognition methods.

Background technology

With economic prosperity and development, the circulation of bank note is increasing, and the multispectral point that domestic many banks use Paper money machine, cleaning-sorting machine, the core technology of ATM are from foreign countries, and not only expensive and presence jeopardizes the hidden danger of financial security.

The core technology basis that multi-optical spectrum paper money counting machine, cleaning-sorting machine, ATM are wanted is real-time banknote image processing and recognized. The country, the image denomination identification technology that multi-optical spectrum paper money counting machine is used is how local using manual features, due to being difficult to avoid that image The problem of skew and taken false distinguishing feature are limited, its stability and recognition capability are all difficult to reach requirement.

Because bank note is that other technical difficult points are：Recognition speed requires high, characteristics of image is difficult to extract, and is improving While recognition capability, it is necessary to ensure robustness of the algorithm to banknote image.Therefore good stability, recognition efficiency are worked out Rmb paper currency recognition methods that is high, can carrying out global recognition is necessary.

The content of the invention

The invention provides a kind of rmb paper currency denomination identifying method of Histogram Matching based on image, it is not according to With the matching of three Color Histograms of denomination banknote image, the denomination of bank note is recognized by the measurement of characteristic vector, is ensured with this Banknote denomination recognition speed improves robustness of the image to noise while high.

The rmb paper currency denomination method for the Histogram Matching based on image that the present invention is provided, its technical scheme is as follows：

The rmb paper currency denomination method of this Histogram Matching based on image, is swept by the high speed of multi-optical spectrum paper money counting machine Retouch, obtain the common N comprising every kind of same denominations image image training sample, by carrying out intermediate value to every training sample Filtering, to reduce the noise from scanning means in image acquisition process, then, to the red, green, blue of every pictures of acquisition Three colouring components extract its histogram, the master die for the banknote image histogram of same denomination being taken average obtained as training Plate；Finally, the sample of the bank note of denomination to be identified is matched with standard form, identifies the real of bank note sample to be identified Denomination；Comprise the following steps that：

First, the dependent image data of each training sample denominations is obtained

S11 obtains the two-dimensional matrix of the image training sample of multiple this kind of training sample denominations；

S12 is filtered to every training sample described image training sample；

S13 extracts the histogram vectors of the colouring component of red, green, blue three of every training sample described image training sample, bag Include the first histogram vectors of red component H_r, the first histogram vectors of green component H_g, and the histogram vectors of blue component first H_b；

S14 obtains the histogrammic average value of each color component in the N training sample images as this kind respectively The standard form of denominations, including red component standard form T (H_r), green component standard form T (H_g), and blueness point Measure standard form T (H_b), wherein；

2nd, denominations image real time transfer to be identified is carried out

S21 obtains the two-dimensional matrix of the images to be recognized of the denominations to be identified；

S22 is filtered to the denominations image to be identified；

S23 extracts the histogram of the redgreenblue component of the images to be recognized, including red component second party figure to Measure h_r, the second histogram vectors of green component h_g, and the second histogram vectors of blue component h_b；

3rd, the real denomination of the denominations to be identified is confirmed

Obtain respectively the distance between second histogram and each denominations template (distance)：

And enter the distance between second Histogram distance of acquisition and each denomination training sample bank note template Row is compared, and is to be determined as the bank note to be identified apart from the minimum corresponding denomination of denominations template of second histogram Denomination.

Applied to banknote denomination identification, banknote denomination said herein includes：100 yuan, 50 yuan, 20 yuan, 10 yuan, 5 yuan, and 1 yuan.

Preferably, in step S11 and step S21, obtained respectively using the high-speed scanning device in multi-optical spectrum paper money counting machine The two-dimensional matrix of image training sample and the two-dimensional matrix of images to be recognized.

Preferably, in step S12 and step S22, intermediate value filter is carried out to image training sample and images to be recognized respectively Ripple.

Preferably, in step S12 and step S22,3 × 3 windows are respectively adopted to image training sample and figure to be identified As carrying out medium filtering.

Preferably, in step s 24, specifically include：

Judged therewith according to second histogram integer value corresponding with the lowest distance value of the denominations template The denomination of the corresponding bank note to be identified.

Histogram matching provided by the present invention for recognizing rmb paper currency denomination, can at least bring following beneficial Effect：

According to the intrinsic image attributes of the bank note of different denominations, i.e., histogrammic distribution is different to be carried out to the denomination of bank note Identification, efficiently solves the drawbacks of traditional method that banknote denomination is recognized by selected characteristic is brought, compared with traditional The recognition methods of banknote denomination, the method that the present invention is provided substantially increases accuracy of identification while ensure that to noise robustness And recognition time is short, the demand that the finance devices such as paper money counter, cleaning-sorting machine, ATM quickly run identification is met well.

Brief description of the drawings

Fig. 1 is the control flow of the rmb paper currency denomination identifying method of the Histogram Matching based on image in the present invention Figure；

Fig. 2 for the present invention in obtain each denominations template specific steps flow chart；

Fig. 3 is the window schematic diagram of medium filtering 3 × 3 in the present invention；

Fig. 4 a are the standard form schematic diagram of 100 yuans of red components in the present invention；

Fig. 4 b are the standard form schematic diagram of 100 yuans of green components in the present invention；

Fig. 4 c are the standard form schematic diagram of 100 yuans of blue components in the present invention.

Embodiment

The present invention is described in further detail with reference to the accompanying drawings and detailed description：

In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below in conjunction with the accompanying drawings and implement The present invention is specifically described example.Drawings in the following description are only some embodiments of the present invention.For this area For those of ordinary skill, on the premise of not paying creative work, other accompanying drawings can also be obtained according to these accompanying drawings.

As shown in figure 1, the invention provides a kind of rmb paper currency denomination identifying side of the Histogram Matching based on image Method, based on principle be being made up of three kinds of basic colors components of RGB for image, treated as long as we extract each banknote The histogram vectors of three kinds of basic colors components of bank note are recognized, and are matched with trained bank note template, you can are known Do not go out the denomination of bank note to be identified.It is worth noting that, the present invention is applied to all current paper moneys in circulation on the market, wherein, RMB includes 100 yuan, 50 yuan, 20 yuan, 10 yuan, 5 yuan and 1 yuan, as long as there is demand in actual applications, is equally applicable to foreign country The identification of the bank note of different denominations, herein we it is not specifically limited.

In the method for the Histogram Matching based on image that the present invention is provided, we are needed to waiting to train in training storehouse first Banknotes of different denominations image be trained, to be compared as template and the image of bank note to be identified.Specific steps bag Include：

As shown in Fig. 2 obtaining the specific steps of the template of each denominations includes：

Specifically, herein, we can obtain N images by the high-speed scanning device in multi-optical spectrum paper money counting machine Training sample.In a specific embodiment, the bank note of different denominations can be directed to, a number of image training is selected respectively Sample, e.g., the image training sample of every kind of banknote denomination choose 10, to ensure the precision for subsequently recognizing bank note to be identified, when So, our quantity to image training sample are not especially limited, although it is more that the quantity of image training sample is chosen, identification Accuracy it is higher, but mean that the time for being trained sample can also lengthen, thus can be selected according to the actual requirements Fixed, such as the image training sample of every kind of banknote denomination chooses 5,15,20,25 etc., or even in the feelings not high to precise requirements Under condition, a training sample as image is only chosen for the bank note of a certain denomination, as long as the mesh of current application can be reached , it is included in present disclosure.

S12 is filtered to every image training sample；

Acquire after a number of image training sample, we are filtered to every training sample, in this hair In bright, in order to improve the robustness to noise, medium filtering is carried out to every image training sample, specifically, in the present invention, As shown in figure 3, medium filtering is carried out to image training sample image using 3 × 3 windows, wherein, p₀,p₁,p₂,p₃,p₄,p₅,p₆, p₇, and p₈Represent respectively to should filter window pixel value and arranged according to size, then P₄=med { P₀,P₁,P₂,P₃, P₄,P₅,P₆,P₇,P₈Be medium filtering result.Obtained respectively after medium filtering medium filtering in red component image, The medium filtering in medium filtering, blue component image in green component image.

S13 extracts the first histogram vectors of the redgreenblue component of every image training sample, including red component First histogram vectors H_r, the first histogram vectors of green component H_g, and the first histogram vectors of blue component H_b；

S14 obtains the histogrammic average value of each color component in N image training samples as this kind of denomination respectively The standard form of bank note, including red component template T (H_r), green component template T (H_g), and blue component template T (H_b), Wherein；

Wherein, N represents the quantity of every kind of banknote image training sample, H_riFor i-th RMB red component histogram；Its Middle H_giFor i-th RMB green component histogram；Wherein H_biFor i-th RMB blue component histogram.

The training process to training sample image is completed by above-mentioned steps, and obtains the denomination of every kind of denominations Bank note template.It is respectively 100 yuans of red component module maps obtained by training as shown in Fig. 4 a, Fig. 4 b and Fig. 4 c (abscissa i represents the scope 0-255 of the pixel value of image), green component Prototype drawing and blue component Prototype drawing (abscissa i tables The scope 0-255 of the pixel value of diagram picture), as 100 yuans of benchmark of identification, (abscissa i represents the pixel value of image Scope 0-255).

Carrying out the specific steps of denominations image real time transfer to be identified includes：

S21 obtains the expression and training of the two-dimensional matrix, here two-dimensional matrix of the images to be recognized of denominations to be identified Sample image obtains identical in template procedure, and therefore not to repeat here.

S22 is filtered to images to be recognized.Similar, here, we are also that it is waited to know using the method for medium filtering Other image is filtered, and therefore not to repeat here.

S23 extracts the histogram of the redgreenblue component of images to be recognized, including red component histogram vectors h_r, it is green Colouring component histogram vectors h_g, and blue component histogram vectors h_b。

Confirm that the real denomination of the denominations to be identified is comprised the following steps that：

Calculate the distance between images to be recognized histogram and each denominations template (distance), that is, treat respectively Recognize the Euclidean distance of the Color Histogram of sample image three and the Color Histogram of template samples three (scope of pixel is 0-255)：

Wherein：

J is image pixel size scope [0,255]；

h_r(j) it is the quantity of j-th of pixel value of banknote image red component histogram to be identified；

T(H_r) (j) for training template image red component j-th of pixel value of histogram quantity；

h_g(j) it is the quantity of j-th of pixel value of banknote image green component histogram to be identified；

T(H_g) (j) for training template image green component j-th of pixel value of histogram quantity；

h_b(j) it is the quantity of j-th of pixel value of banknote image blue component histogram to be identified；

T(H_b) (j) for training template image blue component j-th of pixel value of histogram quantity；

M is that the denomination that has had of whole RMB includes 100 yuan, 50 yuan, 20 yuan, 10 yuan, 5 yuan, and 1 yuan；

K is the size of identification denomination；

By the distance between each denominations of images to be recognized Histogram distance and training sample denominations template It is compared, the minimum corresponding denomination of denominations template of distance is the denomination for being determined as bank note to be identified.

In this step, it is noted that, in a particular embodiment, before the denomination of bank note to be identified is recognized, I Be numbered from big to small according to banknote denomination first, such as：100 yuan of correspondence serial numbers, 1,50 yuan of correspondence serial numbers 2, Qi Tayi Secondary to analogize, M represents the species of denomination.Thus during identification, when we obtain each face of the second Histogram distance During minimum range between volume bank note template, the corresponding integer value of the minimum range judges the face of corresponding bank note to be identified Volume.

The specific embodiment of invention is described in detail above, but the present invention be not restricted to it is described above specific Embodiment, it is intended only as example.To those skilled in the art, any equivalent modifications and replacement carried out to the system Also all among scope of the invention.Therefore, impartial conversion and modification made under the spirit and scope for not departing from invention, It all should be contained within the scope of the invention.

Claims (6)

1. a kind of rmb paper currency denomination identifying method of Histogram Matching based on image, is recognized applied to banknote denomination, its It is characterised by：By the high-velocity scanning of multi-optical spectrum paper money counting machine, the common N comprising every kind of same denominations image image instruction is obtained Practice sample, by being filtered to every training sample, to reduce the noise from scanning means in image acquisition process；So Afterwards, its histogram is extracted to the colouring component of red, green, blue three of every pictures of acquisition, by the banknote image histogram of same denomination The standard form for taking average to be obtained as training；Finally, the sample of the bank note of denomination to be identified is matched with standard form, Identify the real denomination of bank note sample to be identified；Comprise the following steps that：

First, the dependent image data of each training sample denominations is obtained

S11 obtains the two-dimensional matrix of the image training sample of N this kind of training sample denominations；

S12 is filtered to every training sample described image training sample；

S13 extracts the histogram vectors of the colouring component of red, green, blue three of every training sample described image training sample, including red The first histogram vectors of colouring component H_r, the first histogram vectors of green component H_g, and the first histogram vectors of blue component H_b；

S14 obtains the histogrammic average value of each color component in the N training sample images as this kind of denomination respectively The standard form of bank note, including red component standard form T (H_r), green component standard form T (H_g), and blue component mark Quasi-mode plate T (H_b)；

$T\left(H_r\right)=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{H}_{r_i},$

$T\left(H_g\right)=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{H}_{g_i},$

$T\left(H_b\right)=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{H}_{b_i};$

Wherein：N represents the quantity of training sample；H_riFor i-th RMB red component histogram；Wherein H_giFor i-th people Coin green component histogram；Wherein H_biFor i-th RMB blue component histogram；

2nd, denominations image real time transfer to be identified is carried out

S21 obtains the two-dimensional matrix of the images to be recognized of the denominations to be identified；

S22 is filtered to the denominations image to be identified；

S23 extracts the second histogram of the redgreenblue component of the images to be recognized, including the histogram of red component second Vectorial h_r, the second histogram vectors of green component h_g, and the second histogram vectors of blue component h_g；

3rd, the real denomination of the denominations to be identified is confirmed

Obtain respectively the distance between second histogram and each denominations template (distance)：

$dis\tan ce=\underset{1\≤k\≤M}{\min }\sqrt{\frac{1}{3}\underset{j=0}{\overset{255}{\Σ}}{\left(h_r\right(j)-T(H_r\left)\right(j\left)\right)}^2+\frac{1}{3}\underset{j=0}{\overset{255}{\Σ}}{\left(h_g\right(j)-T(H_g\left)\right(j\left)\right)}^2+\frac{1}{3}\underset{j=0}{\overset{255}{\Σ}}{\left(h_b\right(j)-T(H_b\left)\right(j\left)\right)}^2};$

Wherein：J is image pixel size scope [0,255]；

h_r(j) it is the quantity of j-th of pixel value of banknote image red component histogram to be identified；

T(H_r) (j) for training template image red component j-th of pixel value of histogram quantity；

h_g(j) it is the quantity of j-th of pixel value of banknote image green component histogram to be identified；

T(H_g) (j) for training template image green component j-th of pixel value of histogram quantity；

h_b(j) it is the quantity of j-th of pixel value of banknote image blue component histogram to be identified；

T(H_b) (j) for training template image blue component j-th of pixel value of histogram quantity；

M is that the denomination that has had of whole RMB includes 100 yuan, 50 yuan, 20 yuan, 10 yuan, 5 yuan, and 1 yuan；

K is the size of identification denomination；

And compared the distance between second Histogram distance of acquisition and each denomination training sample bank note template It is right, it is the face for being determined as the bank note to be identified apart from the minimum corresponding denomination of denominations template of second histogram Volume.

2. a kind of rmb paper currency denomination identifying method of Histogram Matching based on image as claimed in claim 1, it is special Levy and be, the banknote denomination includes：100 yuan, 50 yuan, 20 yuan, 10 yuan, 5 yuan, and 1 yuan.

3. a kind of rmb paper currency denomination identifying method of Histogram Matching based on image as claimed in claim 1, it is special Levy and be：In step S11 and step S21, image is obtained respectively using the high-speed scanning device in multi-optical spectrum paper money counting machine and is trained The two-dimensional matrix of sample and the two-dimensional matrix of images to be recognized.

4. a kind of rmb paper currency denomination identifying method of Histogram Matching based on image as claimed in claim 1, it is special Levy and be：In step S12 and step S22, medium filtering is carried out to image training sample and images to be recognized respectively.

5. a kind of rmb paper currency denomination identifying method of Histogram Matching based on image as claimed in claim 4, it is special Levy and be：In step S12 and step S22,3 × 3 windows are respectively adopted in image training sample and images to be recognized progress Value filtering.

6. a kind of rmb paper currency denomination identifying method of Histogram Matching based on image as claimed in claim 1, it is special Levy and be：In step s 24, specifically include：

Judge corresponding therewith according to second histogram integer value corresponding with the lowest distance value of the denominations template The bank note to be identified denomination.