CN101967915B - Control method for safety box with palmprint recognition system - Google Patents

Abstract

The invention discloses a control method of a safe with a palmprint recognition system, comprising the following steps: A1: collecting palmprint images; A2: preprocessing the palmprint images; A3: extracting the preprocessed palmprint images; A4: Match the extracted feature vector with the pre-stored palmprint feature vector of the authorized user. If the match is consistent, an unlock command will be issued. If the match is inconsistent, the unlock command will not be issued. Regardless of whether the match is consistent, the unlock command will be issued The unlocking time and user information are stored in the storage module.

Description

A kind of control method of the safety cabinet with Palm Print Recognition System

Technical field

The present invention relates to safety cabinet technical field, relate in particular to a kind of control method of the safety cabinet with Palm Print Recognition System.

Background technology

People are attempting to find a kind of more safety cabinet of " insurance " all the time, because traditional safe system opens with key and password, and key and password are the least roots of " insurance " of safety cabinet.Want to overcome these fatal drawbacks that key and password bring to traditional insurance cabinet, in today of computing machine and mode identification technology develop rapidly, with the self-contained biological characteristic of the mankind, carrying out opening safety cabinet system is a kind of method that domestic and international expert and scholars are finding.

Biological identification technology claims again biological identification technology, refers to the technology that computing machine utilizes the intrinsic physiological characteristic of human body or behavioural characteristic to carry out personal identification evaluation.Biological identification technology is mainly divided into following 8 classes at present: fingerprint recognition, recognition of face, retina identification, iris recognition, speech recognition, signature identification and the dynamics identification etc. of unblanking.Bio-identification have accuracy of identification high, the feature such as be easy to carry, can not lose, can not forget, can not used or usurp, be to have one of new and high technology of development potentiality this century most.In conventional biological identification technology, palmmprint identification is a kind of new technology just growing up in the recent period, compares with other biological characteristic, and palmmprint has the following advantages: (1) has lifelong unchangeability and uniqueness; (2) location has rotational invariance and uniqueness; (3) minutia in palmmprint and all kinds of line feature all have uniqueness and stability; (4) principal character of palmmprint is obvious, is difficult for by noise; (5) be difficult for copying; (6) due to the resolution of palmprint image can be lowered the requirement, collecting device cost is lower; (7) testee's acceptable degree of palmmprint identification is higher; (8) the hardware standard degree of recognition system is also high.

But there is not so far the application process of relevant palmmprint recognition technology in safety cabinet.

Summary of the invention

Technical matters to be solved by this invention is for the deficiencies in the prior art, and a kind of control method of the safety cabinet with Palm Print Recognition System is provided.

The present invention adopts following technical scheme:

A control method with the safety cabinet of Palm Print Recognition System, comprises the following steps:

A1: gather palmprint image;

A2: described palmprint image is carried out to pre-service;

A3: the proper vector of extracting pretreated palmprint image;

A4: according to the authorized user palm print characteristics Vectors matching of the proper vector of extracting and pre-stored, coupling is unanimously sent unlock instruction, mate the inconsistent described unlock instruction that do not send, whether unanimously no matter mate, all uncaging time and user profile are stored in to described memory module.

Described control method, the pre-service that 2 pairs of described palmprint images of described steps A carry out comprises: the binaryzation of palmprint image, extract edge line, determine angle point, location and normalization.

Described control method, described steps A 3 adopts wavelet transformation to extract palmprint image proper vector.

Described control method, adopts Daubechies and Symlet orthogonal wavelet transformation to extract palmprint image proper vector.

Described control method, described palmprint image characteristic extracting module is carried out the 2-d wavelet decomposition of 4 grades to a width palmprint image.

Palmmprint identification safety cabinet of the present invention, to open with the self-contained biological characteristic palmmprint of human body, with other biological characteristic, compare palmmprint and there is lifelong unchangeability and uniqueness, location has rotational invariance, the minutia of palmmprint and all kinds of line feature have stability, the principal character of palmmprint is obvious, is difficult for by noise.Compare the problem that there will not be key loss and copied with the safety cabinet of other form; Also there will not be and forget Password and the stolen problem of password, safe.Nucleus equipment cost is lower, is easy to produce in enormous quantities.

Accompanying drawing explanation

Fig. 1 2-d wavelet decomposing schematic representation;

The 2-d wavelet of Fig. 2 palmprint image decomposes, and wherein a is palmmprint, and b is the wavelet decomposition figure of a palmmprint;

Fig. 3 palmmprint wavelet character vector curve map, wherein a is the palmmprint sample of palm 1,2,3, and b is the proper vector curve of palmmprint 1, and c is the proper vector curve of palmmprint 2, and d is the proper vector curve of palmmprint 3;

The control structure schematic diagram of Fig. 4 palmmprint safety cabinet of the present invention;

Fig. 5 is palmprint image preprocess method schematic diagram.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.

Embodiment 1

The present embodiment provides a kind of control structure of palmmprint safety cabinet, as shown in Figure 4, comprise palm-print image capture module 10, palmprint image pretreatment module 11, palmprint image characteristic extracting module 12, images match module 13, memory module 16 and the control module 15 of unblanking, described palm-print image capture module 10 is for gathering user's palmprint image, described user's palmprint image carries out pre-service in described palmprint image pretreatment module 11, in described palmprint image characteristic extracting module 12, extract the proper vector of pretreated palmprint image, according to the authorized user palm print characteristics Vectors matching of pre-stored in the proper vector of extracting and memory module, coupling is unanimously sent unlock instruction to the described control module of unblanking, mate the inconsistent described unlock instruction that do not send, whether unanimously no matter mate, all uncaging time and user profile are stored in to described memory module,

For example, palm-print image capture module can be used ccd video camera, by PC or DSP, the image after gathering is carried out to pre-service, carries out proper vector extraction, and then mates with the palmmprint being stored in memory module 16.

The palmmprint that is stored in the authorized user in memory module 16 is set up by system in advance by keeper, is authorized user, and authorized user also can add at any time, revise.

The control mechanism concrete steps of the palmmprint safety cabinet of the present embodiment are as follows:

A1: gather palmprint image;

A2: described palmprint image is carried out to pre-service; Described pretreated method is specially:

(1) binaryzation of palmprint image: the palmprint image after cutting apart is carried out to binaryzation, and palm portion is got 0 (black), and background parts got for 1 (in vain), can obtain complete palm type, as shown in Fig. 5-1.

(2) edge extracting: be exactly the point of emptying palm inside to the palmprint image edge extracting after binaryzation.Its basic thought is: if in image a bit for black, and its 8 neighborhood points are while being all black, illustrate that this point is internal point, by this point deletion (being set to white), otherwise, be frontier point, record its position.All pixels in image are carried out to the extraction (shown in Fig. 5-2) that this operation just can complete palmprint image border.

(3) determine angle point: by the method for curve, on the edge line of palmprint image, find out interested angle point A, B, C (Fig. 5-3), wherein A is the intersection point between the third finger and little finger of toe, B is the intersection point between forefinger and middle finger, and C is the mid point of A, B line.

(4) location and normalization: the mid point O that C point is moved to palmprint image, make it to overlap with O point, O is made as true origin (as shown in Fig. 7-3) conventionally, centered by O, rotate palmprint image, AB forwards vertical direction (vertical direction is set to y axle) to, after mobile and rotation, the palmprint image AB line using is all positioned on same position with vertical direction, at the same area place, palm center, cut out a block size and be 128 * 128 subimage and represent whole palmprint image (as Fig. 5-4), can complete the location normalization of palmprint image.

A3: the proper vector of extracting pretreated palmprint image;

A4: according to the authorized user palm print characteristics Vectors matching of the proper vector of extracting and pre-stored, coupling is unanimously sent unlock instruction, mate the inconsistent described unlock instruction that do not send, whether unanimously no matter mate, all uncaging time and user profile are stored in to described memory module.

Preferably, 2 pairs of described palmprint images of described steps A carry out binaryzation, location, normalization, filtering, thinning processing.

Preferably, described steps A 3 adopts wavelet transformation to extract palmprint image proper vector.

Preferably, adopt Daubechies and Symlet orthogonal wavelet transformation to extract palmprint image proper vector.

Preferably, described palmprint image characteristic extracting module is carried out the 2-d wavelet decomposition of 4 grades to a width palmprint image.

Embodiment 2

The present embodiment provides a kind of palmprint image preprocess method.

Pre-service is the first step of the automatic identifying of palmmprint, and its quality directly affects the effect of palmmprint identification.Conventional preprocess method mainly comprises binaryzation, location (rotation and translation), normalization, filtering, the refinement of palmprint image etc.

With the equipment based on CCD, carry out palm-print image capture.In order to reduce error and the distortion causing due to translation, rotation, distortion in gatherer process, when gathering palmprint image, should make every finger of palm open as far as possible, can extract finger edge like this, and aim at and normalization palmprint image with it.After treatment, each palmmprint block size is 128 * 128 (as shown in Fig. 2 (a)), and skew and rotation be eliminated substantially, thereby is conducive to the identification of palmprint image.

Embodiment 3

The present embodiment provides a kind of palmprint image feature extracting method---and wavelet character extracts.

Palmprint image is a kind of approximate grain periodic pattern, and wherein main line is the darkest several the thickest lines on palm, has three main clues on most of palms; Except main line, on palm, also have a lot of pleat lines, in general these lines are thinner, shallow than main line, and very irregular.In palmprint image, the streakline direction of zones of different and spatial frequency are representing the feature of palmprint image inherence.Feature for palmprint image, the methods experiment that we extract through various features, find because wavelet transformation has the function of hyperchannel, multiresolution, therefore by a kind of method based on wavelet transformation, extract the feature of palmprint image, and then identify and not only saved the pretreated time of image, and the robustness of identification is also higher.

Wavelet transformation theory is further developing of Fourier analysis theories, is that people are for the needs of the profound understanding of characteristics of signals.Wavelet transformation provides an adjustable T/F window, flexible result be exactly we can be under different resolution decomposed signal, the result of translation is exactly that we can organize signal as window using this, observes the part of oneself being concerned about.

Definition: if function Ψ (x) meets the small echo condition that allows,

$C_{\mathrm{\&psi;}}={\&Integral;}_{-\&infin;}^{\&infin;}({\left|\widehat{\mathrm{\&psi;}}\left(\mathrm{\&omega;}\right)\right|}^2/\mathrm{\&omega;})\mathrm{d\&omega;}<\&infin;$

Ψ (x) is called and can allows small echo (integration small echo, base small echo) so.Wherein

it is the Fourier transform of Ψ (t).

The wavelet function system being generated by base small echo can be expressed as:

${\mathrm{\&psi;}}_{b,a}\left(t\right)={\left|\frac{1}{a}\right|}^{\frac{1}{2}}\mathrm{\&psi;}\left(\frac{t-b}{a}\right)$

The wavelet transformation of function f (x) is defined as

$\left(W_{\mathrm{\&psi;}}f\right)(b,a)={\left|a\right|}^{-1/2}{\&Integral;}_{-\&infin;}^{\&infin;}f\left(t\right)\stackrel{\&OverBar;}{\mathrm{\&psi;}}\left(\frac{t-b}{a}\right)\mathrm{dt}$

Center and the window width of wavelet function Ψ (t) time window provide with t* and Δ Ψ respectively, frequency window respectively by ω * and

provide, set

${\mathrm{\&psi;}}_{b,a}\left(t\right)={\left|\frac{1}{a}\right|}^{\frac{1}{2}}\mathrm{\&psi;}\left(\frac{t-b}{a}\right)$

The time window of small echo is [b+at ^*-a Δ _Ψ, b+at ^*+ a Δ _Ψ], the width of time window is 2a Δ _Ψ, frequency window is the width of frequency window is

as can be seen here, for the HFS of detection signal, time window can narrow down, and frequency window can broaden; For the low frequency part of detection signal, time window can broaden, and frequency window can narrow down.Therefore, by wavelet transformation time-frequency local characteristics of reflected signal well.Signal is fastened to do at this function and decompose, just obtained the definition of continuous wavelet transform.

The present embodiment adopts wavelet transformation to extract palmprint image feature, and the two-dimentional biorthogonal wavelet that the palmprint image after original location and normalization is carried out to J rank decomposes, and the decomposition of every grade as shown in Figure 1.

As shown in Figure 2, a width palmprint image is carried out to the 2-d wavelet decomposition of 4 grades.In the decomposition result of Fig. 2: the 1st, the upper right corner is cH ₁, the 1st, the lower right corner is cD ₁, the 1st, the lower left corner is cV ₁, other the like.So far, obtained 3J+1 width subimage.These two kinds of information of streakline direction in palmprint image and spatial frequency can be extracted to the 3J width subimage { cH after wavelet decomposition well _k, cV _k, cD _kin go.Each width subimage is asked to normalized 2 norms, thereby obtain the proper vector that length is 3J

$\left\{{[{f_k}^H,{f_k}^V,{f_k}^D]}_{k=1,...,J}\right\},$

Wherein k=1 ..., J

The length of trying to achieve is that the proper vector of 3J can be understood as palmprint image in the different proportion factor (2 ^k) under condition and the general energy distribution on different directions (H, V, D).Get 3 different palms, each palm has gathered 12 image patterns, has certain displacement and angle difference between sample, and Fig. 3 is in 12 samples of each palm.To these 36 samples, try to achieve the proper vector that 3 length are 12, and the proper vector of 12 palmmprint samples of same palm is depicted as to curve map in a width figure, as shown in Figure 3.As can be seen from Figure 3, proper vector exists obvious otherness (different palm) and enough stability (the different samples of identical palm).

Palmprint image database one used has the palmmprint of 20 different palms, and each palm has 12 samples, comes to 240 palmprint images.Palmprint image is that size is 256 grades of gray level images of 128 * 128 pixels, by palm-print image capture module, is collected.Under different small echos and decomposed class (J) condition, test, what discovery effect was best is Daubechies and Symlet orthogonal wavelet, and decomposed class is preferably 4 grades.

Embodiment 4

The present embodiment provides project testing, in project testing, has completed the learning training of 5 people's 50 width (everyone 10 width) palmprint images, then inputting 200 images identifies, everyone 20 width of 5 people in storehouse, stranger's palmprint image 100 width of choosing at random, recognition result is in Table 1.

Table 1 test result

From table 1, the orthogonal wavelet adopting in project has all obtained very high classification accuracy rate.When the sample of mis-classification is analyzed, find that these samples are all palm and collector excessive extrusion owing to gathering, cause streakline to be too out of shape or become close, thereby having changed direction and the spatial frequency feature of palmmprint streakline.

Test result shows, the recognizer of wavelet transformation is very effective.Meanwhile, the key-free safe system that the present invention proposes, if further commercialization can be used as the renewal product of current safe system completely.

Should be understood that, for those of ordinary skills, can be improved according to the above description or convert, and all these improvement and conversion all should belong to the protection domain of claims of the present invention.

Claims (1)

1. a control method with a safe with palmprint recognition system, is characterized in that, comprises the following steps: A1: Collect palmprint images; A2: Preprocessing the palmprint image; the method of the preprocessing is specifically: A21, binarization of palmprint image: carry out binarization to the divided palmprint image, that is, take 0 for the palm part, and take 1 for the background part, and a complete palm shape can be obtained; A22, edge extraction: extracting the edge of the binarized palmprint image is to hollow out the points inside the palm; the basic idea is: if a point in the image is black, and its 8 neighbor points are all black, Explain that this point is an internal point, delete this point, that is, set it as white, otherwise, it is a boundary point, record its position; perform this operation on all pixels in the image to complete the extraction of the palmprint image boundary; A23. Determine the corner points: use the method of curve fitting to find the corner points A, B, and C of interest on the edge line of the palmprint image, where A is the intersection point between the ring finger and the little finger, and B is the point between the index finger and the middle finger C is the midpoint of the line connecting A and B; A24. Positioning and normalization: move point C to the midpoint O of the palmprint image so that it coincides with point O. O is usually set as the coordinate origin, rotate the palmprint image with O as the center, and turn AB to vertical Direction, after moving and rotating, the lines AB and AB of the palmprint image used are positioned at the same position in the vertical direction, and a sub-image with a size of 128×128 is cut out at the same area of the palm center to represent the entire palm. The palmprint image can be used to complete the positioning normalization of the palmprint image; A3: Extract the feature vector of the preprocessed palmprint image; perform 4-level two-dimensional wavelet decomposition on a palmprint image; the first block in the upper right corner is cH ₁ , the first block in the lower right corner is cD ₁ , and the first block in the lower left corner 1 block is cV ₁ , and others are deduced accordingly; so far, 3J+1 sub-images have been obtained; the two kinds of information, ridge direction and spatial frequency in the palmprint image, can be well extracted to the 3J sub-image after wavelet decomposition Go to the image {cH _k , cV _k , cD _k }; find the normalized 2-norm for each sub-image, so as to obtain a feature vector with a length of 3J {[f _k ^H , f _k ^V , f _k ^D ]k=1,...,J}, in $\left\{\begin{array}{c}{f_k}^h={\left|\right|c{h}_k\left|\right|}_2/\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}({\left|\right|c{h}_j\left|\right|}_2+{\left|\right|c{V}_j\left|\right|}_2+{\left|\right|c{\mathrm{D.}}_j\left|\right|}_2)\\ {f_k}^V={\left|\right|c{V}_k\left|\right|}_2/\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}({\left|\right|c{h}_j\left|\right|}_2+{\left|\right|c{V}_j\left|\right|}_2+{\left|\right|c{\mathrm{D.}}_j\left|\right|}_2)\\ {f_k}^{\mathrm{D.}}={\left|\right|c{\mathrm{D.}}_k\left|\right|}_2/\underset{j=1}{\overset{J}{\mathrm{\&Sigma;}}}({\left|\right|c{h}_j\left|\right|}_2+{\left|\right|c{V}_j\left|\right|}_2+{\left|\right|c{\mathrm{D.}}_j\left|\right|}_2)\end{array}\right.$ k=1,...,J The obtained feature vector with a length of 3J can be understood as the approximate energy distribution of the palmprint image under different scale factors and in different directions; 3 different palms are taken, and 12 image samples are collected for each palm. samples, obtain 3 feature vectors with a length of 12, and draw the feature vectors of 12 palmprint samples of the same palm into a graph; A4: Match the extracted feature vector with the pre-stored palmprint feature vector of the authorized user. If the match is consistent, the unlock command will be issued. If the match is inconsistent, the unlock command will not be issued. No matter whether the match is consistent, the unlock time and user information will be stored. in the storage module.

Images