Background technology
In modern society, sharply frequent along with the high development of network technology, flow of personnel, a kind of identity authorization system safe and reliable, convenience and high-efficiency seems particularly important.Traditional identification mainly contains two kinds: mark article (key, identity document etc.) and mark knowledge (user name, password etc.).But in actual applications,, there is easily the deficiencies such as loss property, easily forgery property, nonuniqueness and range of application be relatively little in it, makes people in the urgent need to a kind of personal identification method that can overcome above-mentioned defect.Under the driving of demand, the recognition technology based on biological characteristics such as face, fingerprint, iris, hand shape, person's handwritings is arisen at the historic moment.
At present, there are a lot of iris authentication systems both at home and abroad.The iris authentication system based on Gabor that wherein doctor Daugman realizes be propose the earliest can practical application iris authentication system.In iris authentication system, first to gather iris image, then in the iris image collecting, cut apart iris region, finally after normalization, on iris image, extract feature and mate.Wherein iris feature extraction is one of principal element affecting system performance, and the quality requirements of the large multipair iris image of existing Algorithm of Iris Recognition is higher.But, the impact of the factor such as the iris image collecting under natural light can be subject to eyelashes, eyelid, illumination, rock, cause the iris image quality collecting not good enough, for this class inferior quality iris image, need a kind of effectively iris feature to extract and recognizer.
Summary of the invention
The object of this invention is to provide one based on multi-direction Gabor and Adaboost iris identification method, there is good recognition performance, overcome the deficiency of the above-mentioned aspect of prior art.
The object of the invention is to be achieved through the following technical solutions:
A kind of based on multi-direction Gabor and Adaboost iris identification method, it comprises the steps:
1) to normalized iris image piecemeal and extract two-dimensional Gabor feature, and calculate Hamming distance between corresponding blocks from, it specifically comprises the following steps:
1.1) iris image of expansion is evenly divided into M is capable, N row, obtain M × N iris image submodule;
1.2) use the Gabor wave filter of eight directions of same yardstick to act on step 1.1) in iris image submodule after normalization, then according to Gabor real part, the positive and negative of image filtering result encoded; Wherein the expression formula of Gabor wave filter is as follows:
,
,
,
the direction of Gabor wave filter,
, ,
uwith
vbe respectively the horizontal and vertical centre frequency of Gabor wave filter,
with
respectively Gaussian envelope along
axle and
the space constant of axle, it represents the yardstick of Gabor wave filter;
The expression formula of feature coding is as follows:
Wherein
iiris image,
ggabor wave filter,
represent the real imaginary part of symbol of filtering result,
for feature coding;
1.3) according to formula:
, calculate need M × N iris image submodule corresponding to two iris images of coupling Hamming distance separately from, for each iris image submodule, can obtain eight Hamming distances from, eight Hamming distances are designated as to V from the proper vector of formation
whole,
;
Wherein code A and code B represent respectively the Gabor feature coding of two iris images; Mask A and mask B represent respectively the noise template of two iris images, and its value, for the effective iris portion of " 1 " interval scale, is " 0 " interval scale noise;
1.4) make L=M × N, according to step 1.3) in Hamming distance obtain 8 × L Hamming distance from V from computing formula
part, be designated as:
1.5) by the Hamming distance of whole iris image from V
partwith the Hamming distance of iris image submodule from V
wholemerge, form a proper vector V that dimension is 8+8 × M × N, V=[V
whole,v
part], that is:
2) use Adaboost algorithm to carry out Classification and Identification to the piecemeal distance obtaining in step 1), it specifically comprises the following steps:
2.1) a sample training collection is set
,
,
, i=1,2 ..., N, M is vector
dimension, wherein:
the multidimensional characteristic vectors that to be the proper vector that forms of whole iris image be combined into piecemeal proper vector afterwards, when two iris image submodules are during from same human eye,
, otherwise
;
2.2) pass through formula:
, sample training collection is carried out to weight and go out beginningization, then the sample training collection that has weight distribution is carried out to training study, obtain a Weak Classifier
;
; T=1,2 ..., T, T is iterations;
2.3) select a feature in step 1), then pass through formula:
, m=1,2 ..., M, calculates Weak Classifier
in the concentrated classification error rate of sample training
, note
, when
time, make T=t-1, and jump out circulation;
2.4) pass through formula:
, calculate Weak Classifier
weight
;
2.5) pass through formula:
, upgrade sample training centralization of state power weight, and sample training collection be normalized;
2.6) finally according to formula:
, obtaining final sorter, the feature of this sorter is the feature of proper vector V in iris image, and identification completes.
Beneficial effect of the present invention is: only use a less part or the multiple part of Noise in iris to identify, reduced the impact of noise, well solved the identification problem of inferior quality iris image, have good recognition performance.
Embodiment
As shown in Figure 1, the one described in the present embodiment is based on multi-direction Gabor and Adaboost iris identification method, and it comprises the steps:
Step 1: normalized iris image is extracted to two-dimensional Gabor feature
2D Gabor wave filter is owing to having good resolution characteristic in time domain and frequency domain, and expression formula is as follows:
Wherein
,
,
the direction of Gabor wave filter,
with
be respectively the horizontal and vertical centre frequency of Gabor wave filter,
with
respectively Gaussian envelope along
axle and
the space constant of axle, represents the yardstick of Gabor wave filter.
The present invention uses the Gabor wave filter (Fig. 2) of eight directions of same yardstick, corresponding respectively
,
.
These eight two-dimensional Gabor filter are acted on to the iris image after normalization, use the quadrant of filtering result to encode, be expressed as follows:
Wherein
iris image,
gabor wave filter,
represent the real imaginary part of symbol of filtering result,
for feature coding.
Whole iris region has comprised the noise such as hot spot, eyelashes, therefore, in the time only using certain part of few Noise in iris to identify, can reduce to a certain extent the impact of noise, likely improves on the contrary recognition performance.As Fig. 4, by the iris image of expansion be evenly divided into that M is capable, N row, so just obtain
individual sub-block.To these sub-block iris implementation steps 1.
The present invention adopts Hamming distance from as similarity measurement, as Fig. 3, the Gabor coding of two width images is carried out to XOR, consider the noise in two width images, using the XOR result of effective pixel points add and divided by the number of effective pixel points as an eigenwert, be Hamming distance from, formula is as follows:
Wherein
with
it is respectively the Gabor feature coding of iris image A and iris image B;
with
represent respectively the noise template of iris image A and iris image B, its value, for the effective iris portion of " 1 " interval scale, is " 0 " interval scale noise.
First, calculate whole iris unfolded image Gabor coding Hamming distance from, because the present invention has chosen the Gabor wave filter of eight directions, so can obtain eight Hamming distances from the proper vector forming, be designated as:
Then, calculate and need two irises of coupling corresponding
individual piecemeal Hamming distance separately from, for each sub-block, can obtain eight Hamming distances from, order
, just can obtain
individual
Hamming distance from, be designated as:
Finally, will
with
merge, obtain proper vector
, that is:
(6)
The dimension of proper vector is
.
Step 2: use Adaboost algorithm to carry out Classification and Identification to feature
Adaboost algorithm is the simple classification device (being called Weak Classifier) that utilizes a large amount of classification capacities general, combines the sorter that composition and classification is very capable by certain method.Detailed process is as follows:
Input: training set
, wherein
,
,
,
it is vector
dimension; Iterations T and weak learning algorithm.
Initialization: weight
.(7)
1) to there being the training set study of weight distribution, obtain a Weak Classifier
2) select best feature to make classification error rate
minimum, wherein
(9)
, and note
.If
, order
and jump out circulation.
3) calculate Weak Classifier
weight:
4) upgrade sample weights:
Wherein
it is normalized factor.
Iris recognition problem is a typical classification problem.As long as make each Weak Classifier corresponding to 1 feature (being also Hamming), and according to the judgement of classifying of the size of eigenwert, iris recognition process is exactly the process that Adaboost selects feature, namely selects the process of Weak Classifier.In the process of classifying at two width images, use
As feature, in training set
; Take two images that obtain for identical human eye, order
, different human eyes are taken two images that obtain, order
.Then adopt AdaBoost algorithm to train, the characteristic of division of choosing, then these characteristic of divisions are combined into stronger sorter.
The present invention is not limited to above-mentioned preferred forms; anyone can draw other various forms of products under enlightenment of the present invention; no matter but do any variation in its shape or structure; every have identical with a application or akin technical scheme, within all dropping on protection scope of the present invention.