CN104268529A - Judgment method and device for quality of fingerprint images - Google Patents
Judgment method and device for quality of fingerprint images Download PDFInfo
- Publication number
- CN104268529A CN104268529A CN201410510227.0A CN201410510227A CN104268529A CN 104268529 A CN104268529 A CN 104268529A CN 201410510227 A CN201410510227 A CN 201410510227A CN 104268529 A CN104268529 A CN 104268529A
- Authority
- CN
- China
- Prior art keywords
- fingerprint image
- quality
- optimal classification
- hog
- classification surface
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000004364 calculation method Methods 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 15
- 239000013598 vector Substances 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 9
- 238000012706 support-vector machine Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 abstract description 5
- 238000002474 experimental method Methods 0.000 abstract description 4
- 239000007787 solid Substances 0.000 abstract 1
- 230000006870 function Effects 0.000 description 12
- 238000001514 detection method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 210000004243 sweat Anatomy 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000009394 selective breeding Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
- G06V40/1365—Matching; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Abstract
The invention discloses a judgment method and device for the quality of fingerprint images. The judgment method includes the step of obtaining fingerprint image samples, the step of carrying out learning according to the fingerprint image samples by using an SVM classifier to obtain the optimal classification face, the step of obtaining the images to be judged and calculating histogram of oriented gradient (HOG) characteristics of the fingerprint images, and the step of judging the quality of the fingerprint images according to the HOG characteristics and the optimal classification face. Thus, the quality of the fingerprint images is accurately judged in small size fingerprint acquisition equipment, the work of manually determining and judging threshold values is omitted, good expansion capability is achieved, influences brought by various types of noise can be determined in the judgment mode, judgment can be completed just by inputting required sample types, an excellent effect is exerted in a lot of experiments, and a solid foundation is laid for improving the fingerprint recognition rate and lowering the false acceptance rate.
Description
Technical field
The present invention relates to communication technical field, especially relate to a kind of determination methods and device of fingerprint image quality.
Background technology
Fingerprint identification technology has started to be widely used in mobile terminal, and for the fingerprint recognition of mobile terminal, recognizer is then core technology.Due to the restriction of fingerprint collecting sensor, when finger is with in the situations such as sweat stain, mud stain, discrimination can reduce greatly, particularly this small size fingerprint collecting of mobile terminal equipment, recognizer relies on the fingerprint image quality of input especially, the higher then discrimination of quality is higher, and accuracy of system identification is lower, therefore particularly important to the judgement of fingerprint image quality before recognition.
Traditional determination methods, first adds up the characteristic indexs such as the average of fingerprint image, variance, information entropy, annular spectrum structure, then these index calculate quality score comprehensive.But these class methods are only applicable to entirely gather fingerprint recognition system, and inapplicable for small size fingerprint acquisition system; Moreover these characteristic indexs do not take into full account that fingerprint is the image with particular texture structure yet.Therefore, the determination methods of fingerprint image quality of the prior art, can't judge accurately to fingerprint image quality in small size fingerprint collecting equipment (or system).
Summary of the invention
Fundamental purpose of the present invention is the determination methods and the device that provide a kind of fingerprint image quality, is intended to automatic decision fingerprint image quality, and for improving fingerprint recognition rate, reduction accuracy of system identification is laid a good foundation.
To achieve these objectives, the present invention proposes a kind of determination methods of fingerprint image quality, comprises step:
Obtain fingerprint image sample;
Utilize SVM support vector machine classifier to learn according to described fingerprint image sample, obtain optimal classification surface;
Obtain fingerprint image to be judged, and calculate the HOG gradient orientation histogram feature of described fingerprint image;
The quality of described fingerprint image is judged according to described HOG characteristic sum optimal classification surface.
Preferably, the described SVM of utilization support vector machine classifier according to described fingerprint image sample carry out study obtain optimal classification surface comprise:
Calculate the HOG feature of described fingerprint image sample;
The HOG feature of described fingerprint image sample is inputted in described SVM classifier and carries out training study, obtain optimal classification surface.
Preferably, the HOG gradient orientation histogram feature of the described fingerprint image of described calculating comprises:
Calculate the gradient direction value of each location of pixels in described fingerprint image;
Described fingerprint image is divided into multiple cell, is that each cell builds gradient orientation histogram according to described gradient direction value;
Described cell is combined into block, gradient orientation histogram described in normalization in described piece, the gradient orientation histogram of all pieces in described fingerprint image is combined and forms HOG feature.
Preferably, describedly judge that the quality of described fingerprint image comprises according to described HOG characteristic sum optimal classification surface:
The HOG feature of described fingerprint image inputted in function f (the x)=w*x-b of described optimal classification surface and calculate, wherein w is the support vector of optimal classification surface, and b is constant term, and x is the HOG feature of fingerprint image;
If result of calculation is f (x) > 0, then judge that the quality of described fingerprint image is good;
If result of calculation is f (x) < 0, then judge the of poor quality of described fingerprint image.
The present invention proposes a kind of judgment means of fingerprint image quality simultaneously, comprises study module and judge module, wherein:
Study module, for obtaining fingerprint image sample, utilizes SVM classifier to learn according to described fingerprint image sample, obtains optimal classification surface;
Judge module, for obtaining fingerprint image to be judged, and calculating the HOG gradient orientation histogram feature of described fingerprint image, judging the quality of described fingerprint image according to described HOG characteristic sum optimal classification surface.
Preferably, described study module is used for: the HOG feature calculating described fingerprint image sample, the HOG feature of described fingerprint image sample is inputted in described SVM classifier and carries out training study, obtains optimal classification surface.
Preferably, described judge module comprises processing unit, and described processing unit is used for: the gradient direction value calculating each location of pixels in described fingerprint image; Described fingerprint image is divided into multiple cell, is that each cell builds gradient orientation histogram according to described gradient direction value; Described cell is combined into block, gradient orientation histogram described in normalization in described piece, the gradient orientation histogram of all pieces in described fingerprint image is combined and forms HOG feature.
Preferably, described judge module comprises processing unit and judgement unit, wherein:
Processing unit, for the HOG feature of described fingerprint image is inputted described optimal classification surface function f (x)=w*x-b in calculate, wherein w is the support vector of optimal classification surface, and b is constant term, and x is the HOG feature of fingerprint image;
Judgement unit, differentiates for the result of calculation according to processing unit, if result of calculation is f (x) > 0, then judges that the quality of described fingerprint image is good; If result of calculation is f (x) < 0, then judge the of poor quality of described fingerprint image.
The determination methods of a kind of fingerprint image quality provided by the present invention, by SVM classifier, study is carried out to fingerprint image sample and obtain optimal classification surface, and HOG feature is incorporated in fingerprint image quality judgement, according to the quality of HOG characteristic sum optimal classification surface automatic decision fingerprint image.Not only eliminate the work manually determining judgment threshold, and there is good extended capability, namely this judgment mode can judge the impact that polytype noise brings, and the sample type that only need input needs can complete judgement, shows splendid effect in great many of experiments.Determination methods of the present invention is mainly applicable to the judgement of small size fingerprint collecting equipment to the fingerprint image quality quality gathered, especially for the situation that sweat stain, mud stain, noise etc. cause fingerprint image fuzzy, can judge accurately the quality of fingerprint image before fingerprint recognition, for improving fingerprint recognition rate, reduction accuracy of system identification is laid a good foundation.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of determination methods one embodiment of fingerprint image quality of the present invention;
Fig. 2 is the process flow diagram of the HOG feature calculating fingerprint image sample in the embodiment of the present invention;
Fig. 3 is the schematic diagram of optimal classification surface in the embodiment of the present invention;
Fig. 4 is the process flow diagram of the HOG feature calculating fingerprint image in the embodiment of the present invention;
Fig. 5 is the structured flowchart of judgment means one embodiment of fingerprint image quality of the present invention;
Fig. 6 is the structured flowchart of judge module in the embodiment of the present invention.
The realization of the object of the invention, functional characteristics and advantage will in conjunction with the embodiments, are described further with reference to accompanying drawing.
Embodiment
Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.
The determination methods of fingerprint image quality of the present invention, consider HOG (Histogram of Oriented Gradient, gradient orientation histogram) feature interpretation image gradient directional spreding, it is effective Texture Statistical Feature, therefore by introducing HOG feature and adopting SVM supervised learning pattern successfully to distinguish fingerprint image quality, achieve and fingerprint image quality in small size fingerprint collecting equipment is judged accurately.
See Fig. 1, propose determination methods one embodiment of fingerprint image quality of the present invention, described determination methods comprises the following steps:
Step S101: obtain fingerprint image sample
Fingerprint image sample comprises positive and negative sample, i.e. the measured sample of matter and ropy sample, at least each one of positive negative sample, preferably multiple.Fingerprint image sample is by artificial selection, and can be the on-the-spot artificial fingerprint image gathering acquisition, also can be the ready-made fingerprint image obtained from outside.
Step S102: utilize SVM classifier to learn according to fingerprint image sample, obtains optimal classification surface
Concrete, the first HOG feature of calculated fingerprint image pattern, then carries out training study by the HOG feature of fingerprint image sample input SVM (Support Vector Machine, support vector machine) sorter, thus obtains optimal classification surface.
HOG feature is a kind of Feature Descriptor of object detection, and it carrys out constitutive characteristic by the gradient orientation histogram of statistical picture regional area.Gradient orientation histogram describes image gradient directional spreding, is effective texture statistics characteristic.The computing method of the HOG feature of fingerprint image sample as shown in Figure 2, comprise the following steps:
Step S121: the gradient direction value of each location of pixels in calculated fingerprint image pattern
Concrete, the gradient of transverse and longitudinal coordinate direction in calculated fingerprint image pattern, and root this calculate the gradient direction value of each location of pixels.In fingerprint image sample, the gradient at pixel (x, y) place is:
G
x(x,y)=I(x+1,y)-I(x-1,y)
G
y(x,y)=I(x,y+1)-I(x,y-1)
Wherein G
x(x, y), G
y(x, y) and I (x, y) represent the horizontal direction gradient at pixel (x, y) place in fingerprint image sample, vertical gradient and image intensity value respectively.Gradient magnitude G (x, y) and the gradient direction θ (x, y) at pixel (x, y) place are respectively:
Step S122: fingerprint image sample is divided into multiple cell is that each cell builds gradient orientation histogram according to gradient direction value
Concrete, suppose that each cell (cell) is for 6*6 pixel, adopt the histogram of 9 bin to add up the gradient information of this 6*6 pixel, namely the gradient direction 360 degree of cell is divided into 9 direction blocks.Such as: if the gradient direction value of this pixel is 20-40 degree, gradient magnitude is 2, then the counting of histogram second bin is exactly 2, like this projection (being mapped to fixing angular range) is weighted to pixel gradient orientation histogram each in cell, just can obtain the gradient orientation histogram of this cell, 9 dimensional feature vectors (because having 9 bin) that also namely this cell is corresponding.
Step S123: cell is combined into block, normalized gradient direction histogram in block
Concrete, unit lattice be combined into larger, coconnected interval, space and form block (block), like this, in a block, the proper vector of all cells is together in series and just obtains the HOG feature of this block.These intervals overlap each other, and this just means: the feature of each cell repeatedly can appear in last proper vector with different results.Block descriptor after normalization is just referred to as HOG descriptor by us.
Step S124: the gradient orientation histogram of all pieces in fingerprint image sample is combined and forms HOG feature
Finally overlapping blocks all in fingerprint image sample are carried out HOG feature collection, and they are combined into final proper vector, be i.e. the HOG feature of fingerprint image sample.
In certain embodiments, limit a detection window in the local of fingerprint image sample, only this detection window is carried out to the calculating of HOG feature.Or, progressively carry out scanning the HOG feature obtaining whole fingerprint image sample to whole fingerprint image sample in the mode of detection window.
After calculating the HOG feature of fingerprint image sample, next, SVM classifier then carries out training study according to the HOG feature of multiple fingerprint image sample, obtains optimal classification surface.SVM is the very classical supervised learning method (namely manually determining sample) in machine learning field, has been widely used at present in the fields such as text classification, has achieved good effect.
Finger print image quality analysis can regard two classification problems as, and namely quality is well+1, and of poor quality is-1.For two classification problems, SVM classifier needs in higher dimensional space, find a linear lineoid, will belong to the data point of two classifications separately.For Fig. 3, in two dimensional surface, adularescent and black two class point, need to find a plane by 2 class points separately.By the situation in figure, plane can find unlimited, and SVM classifier is using from two class data point border planes farthest as optimal classification surface, namely makes two class data points to the Maximizing Minimum Distance of plane.
In specific implementation, a lineoid can be defined as optimal classification surface, namely the plane that the solid line in figure represents, it makes the point of label value y=-1 drop on the side of f (x) < 0, and the point of y=+1 drops on the side of f (x) > 0.The function of definition optimal classification surface is f (x)=wx-b, and wherein w is the support vector of optimal classification surface, and b is constant term, and x is the HOG feature of the image of input, w and b is the parameter of optimal classification surface.In addition a support hyperplane is defined again, the plane H1 that namely in figure, two dotted lines represent and H2.Assuming that | (w, b) |=1, then the formula of support hyperplane is wx-b=± 1.According to the hypothesis of SVM, for optimal classification surface, all data points should meet y (wx-b) >=1.Can release, the distance d=2/|w| between two support hyperplanes simultaneously.The objective function of SVM can be obtained thus:
Utilize method of Lagrange multipliers to solve to the problems referred to above, can obtain:
The parameter w of final acquisition optimal classification surface and b.As can be seen from each formula above, the parameter w of optimal classification surface is linearly added by sample point and obtains, and these sample points must drop between two support hyperplanes, and these sample points are called as support vector.And the weight of sample point outside two support hyperplanes is 0, on decision optimal hyperlane without any impact.
Therefore, the optimal classification surface that SVM classifier is tried to achieve is only by the impact of a part of data, performance is just relatively stable, and often has good Generalization Capability by the lineoid that maximize margin is tried to achieve, namely do to unknown data classify in can have lower error rate.
After obtaining optimal classification surface by automatic training study, then can judge the quality of fingerprint image.
Step S103: obtain fingerprint image to be judged
This fingerprint image can be the fingerprint image of collection in worksite, also can be the fingerprint image obtained from outside.
Step S104: the HOG feature of calculated fingerprint image
The concrete calculation process of the HOG feature of fingerprint image as shown in Figure 4, comprises the following steps:
Step S141: the gradient direction value of each location of pixels in calculated fingerprint image
Step S142: fingerprint image is divided into multiple cell is that each cell builds gradient orientation histogram according to gradient direction value
Step S143: cell is combined into block, normalized gradient direction histogram in block
Step S144: the gradient orientation histogram of all pieces in fingerprint image is combined and forms HOG feature
The computing method of the HOG feature of fingerprint image in this step S104, identical with the computing method of the HOG feature of fingerprint image sample in step S102, do not repeat them here.
Step S105: the quality judging fingerprint image according to HOG characteristic sum optimal classification surface
Concrete, calculate in function f (the x)=w*x-b of the HOG feature of fingerprint image input optimal classification surface, wherein w is the support vector of optimal classification surface, and b is constant term, and x is the HOG feature of fingerprint image.If result of calculation is f (x) > 0, then judge that the quality of fingerprint image is good, output label value y=+1; If result of calculation is f (x) < 0, then judge the of poor quality of fingerprint image, output label value y=-1.
For example, suppose to obtain optimal classification surface after learning sample, its parameter constant item b is 2.8326173060507497e+001, and support vector w is as follows, is one 1296 dimension data:
-5.65062046e-001?-1.22261524e-001?6.23945475e-001?2.61808070e-003
-5.96388519e-001?7.09277689e-001?-1.10982299e+000?2.81930566e-001
1.25543416e+000?-2.58267093e+000?-3.93080682e-001?1.26033258e+000
7.85001144e-002?-3.78241152e-001?7.57321477e-001?2.73346394e-001
2.01390719e+000?4.89104331e-001?6.39280975e-001?5.81095874e-001
2.78728080e+000?-2.40852520e-001?-1.01792566e-001?1.00278020e+000
1.44761455e+000?-1.58462083e+000?1.89574575e+000?1.07644367e+000
8.66438568e-001?-6.86612964e-001?9.07730460e-001?-3.49271774e-001
1.22522628e-002?-6.30278111e-001?3.00675720e-001?1.85192859e+000
1.62126279e+000?4.82157320e-001?-1.34164858e+000?2.04924569e-001
1.55832696e+000?1.77290034e+000?-5.20038784e-001?7.18480110e-001
2.09454685e-001?7.86378205e-001?-1.78212130e+000?9.680963?16e-001
8.70740235e-001?-6.44824266e-001?2.54506993e+000?7.10121095e-001
-5.98418236e-001?-1.06928796e-001?8.71971488e-001?1.17480731e+000
7.15802729e-001?2.14615989e+000?-2.88301253e+000?-1.92418182e+000
2.96458960e-001?-8.55592549e-001?-2.03400755e+000?-6.68073654e-001
1.38882890e-001?-7.04979658e-001?-1.46523386e-001?1.70489192e+000
-1.59027326e+000?-1.56655461e-001?-1.12606674e-001?3.60005188e+000
-2.28805041e+000?1.52749741e+000?1.74320495e+000?-6.91233456e-001
2.85730386e+000?-1.05900079e-001?-1.68506277e+000?-1.18375808e-001
-1.10159695e+000?6.55481517e-001?-8.89651656e-001?5.75935364e-001
-8.52293670e-001?-1.37630284e+000?7.58536041e-001?1.52515018e+000
6.06581271e-001?1.50268352e+000?1.77640891e+000?-7.86289200e-002
1.14980662e+000?1.16938233e+000?2.01216507e+000?-6.11184180e-001
-7.25955606e-001?-2.55225211e-001?7.07862496e-001?4.88763779e-001
7.70018041e-001?-1.96791291e+000?2.75504446e+000?-9.32191432e-001
-2.94734448e-001?6.88520610e-001?-5.86638331e-001?1.37036932e+000
-1.00271665e-001?1.41746268e-001?-2.22995734e+000?2.95861661e-001
2.21844387e+000?1.66081011e+000?-4.16236019e+000?4.55661625e-001
1.07015693e+000?2.21201491e+000?-1.26312482e+000?-9.47408438e-001
2.54873562e+000?1.56559813e+000?-1.24778640e+000?-1.32051051e+000
1.65815008e+000?-1.41459537e+000?1.92478895e+000?-1.84008098e+000
3.61277342e+000?2.20739937e+000?-1.34330928e+000?-5.74071646e-001
1.62802529e+000?7.33614326e-001?-2.55355924e-001?-1.70272529e+000
1.97496819e+000?-1.09231484e+000?-2.04441524e+000?-3.81036431e-001
2.97871494e+000?-3.68865132e+000?3.15781379e+000?5.97642064e-001
5.80903649e-001?-2.13103652e+000?-8.67208302e-001?1.80704606e+000
9.69478860e-002?4.28946495e-001?5.80691278e-001?-5.95060766e-001
9.76634204e-001?-7.90543929e-002?-4.92165685e-001?1.70432973e+000
-1.05518484e+000?8.51133049e-001?-1.17313042e-001?1.35035202e-001
2.53540706e-002?6.81154847e-001?-1.51302606e-001?-1.07936570e-002
-1.3?1098163e+000?2.20081162e+000?-3.33757579e-001?1.28789783e+000
-1.38676023e+000?-5.06396294e-001?1.09426820e+000?3.83968383e-001
1.24972117e+000?2.32150960e+000?5.33086717e-001?1.50917560e-001
5.07387817e-001?-1.61430746e-001?2.17941809e+000?3.69923621e-001
2.83043933e+000?1.90740168e-001?1.46296895e+000?1.37301970e+000
6.74435854e-001?-6.71533048e-001?-9.91295278e-001?9.93396699e-001
-3.25666308e-001?1.57782927e-001?-1.24856126e+000?-1.44502115e+000
8.13267171e-001?1.03857972e-001?-2.26438427e+000?1.86400020e+000
1.52461588e+000?2.44749278e-001?1.56721997e+000?-1.32312346e+000
-8.10653090e-001?-2.48049855e+000?2.99090832e-001?5.61489940e-001
-2.98761845e-001?-3.92122388e+000?-1.46208012e+000?-2.30779409e+000
-5.69158137e-001?9.63538647e-001?1.94473159e+000?-8.96440353e-003
1.95122707e+000?-2.46136642e+000?2.12227440e+000?2.57141161e+000
-1.22917616e+000?2.47263566e-001?-1.08239722e+000?4.82539266e-001
-1.97637364e-001?-4.98678893e-001?-7.77484000e-001?1.73539460e+000
4.62011158e-001?6.98073208e-003?7.24058747e-001?-3.12087923e-001
1.10251451e+000?1.25460291e+000?1.75949061e+000?-5.29235363e-001
-9.64799151e-002?-2.53143954e+000?-4.57035750e-001?3.02866429e-001
1.37184167e+000?3.23770791e-001?-2.42519689e+000?5.12698852e-002
-8.26845348e-001?-1.34545541e+000?-8.93717766e-001?1.43910718e+000
9.63344097e-001?-8.71918917e-001?-1.07027996e+000?1.02522659e+000
1.23496354e+000?4.57389086e-001?1.19853270e+000?1.64693880e+000
5.11548460e-001?-1.33672154e+000?2.71858722e-001?-1.17530978e+000
6.15254566e-002?2.94417411e-001?4.36541080e-001?1.37163055e+000
-1.82581842e+000?-6.85078382e-001?-7.76390493e-001?-1.97848654e+000
2.16851807e+000?1.60647643e+000?-1.86727822e-001?6.96594656e-001
-7.91485071e-001?2.65202731e-001?1.50661623e+000?1.07845032e+000
1.78456712e+000?-1.79334183e-003?1.40141740e-001?1.53397667e+000
1.23211336e+000?7.10765898e-001?5.86186089e-002?2.19317412e+000
2.13756412e-001?3.66310835e-001?5.48000991e-001?-6.51535928e-001
-1.14362442e+000?-4.40639943e-001?1.27369392e+000?1.44236636e+000
1.37426305e+000?2.02852416e+000?-3.16964775e-001?-2.62251306e+000
4.59316635e+000?-1.96874535e+000?8.52879345e-001?-2.99564815e+000
7.21852005e-001?9.10931945e-001?-2.10801840e+000?-8.83882880e-001
1.07756746e+000?-1.12351263e+000?3.39177704e+000?1.30952287e+000
3.86621094e+000?3.64660531e-001?1.86861193e+000?1.12214160e+000
1.42694950e-001?-6.56762421e-001?-6.42201722e-001?7.56536007e-001
-7.53396690e-001?1.51442754e+000?-1.55608511e+000?-3.20019364e-001
1.42415535e+000?-1.35575104e+000?7.34291017e-001?1.73662949e+000
-9.04838800e-001?-1.99635935e+000?-1.95133519e+000?-8.14499021e-001
3.27238366e-002?3.87542397e-001?-1.95944226e+000?-1.91607440e+000
2.20693421e+000?1.58080161e+000?-3.00249863e+000?-1.07961583e+000
-6.12612069e-001?-9.43629518e-002?2.07135153e+000?-1.07898617e+000
1.98396698e-001?-8.05510044e-001?9.38360572e-001?-1.17815860e-001
-1.42652404e+000?-1.76703289e-001?-4.94757183e-002?-1.07781088e+000
8.66706133e-001?-1.95579135e+000?-1.15027654e+000?1.17117381e+000
-7.06813395e-001?1.68010008e+000?6.21922314e-001?1.40011585e+000
-5.73635876e-001?5.50615430e-001?3.30518556e+000?3.77227396e-001
7.73974836e-001?3.53976369e+000?1.18661797e+000?1.13880858e-001
1.61683464e+000?3.84422779e-001?-7.32956529e-001?-2.02169132e+000
2.86095762e+000?-5.44128776e-001?1.94903874e+000?3.54368895e-001
9.16079760e-001?1.09807134e+000?-8.51657987e-002?-8.94440413e-002
1.56543106e-001?-7.25803673e-001?-2.16020489e+000?9.41323280e-001
-2.19787073e+000?1.78099692e+000?1.21620715e+000?1.99195230e+000
-1.74088085e+000?-1.65861666e+000?1.76478863e+000?2.54104710e+000
-1.49221981e+000?5.00257432e-001?8.48669946e-001?-1.79053199e+000
-2.45864248e+000?2.42074177e-001?1.22374451e+000?-8.02697659e-001
-1.12856038e-001?-1.45404863e+000?-1.14215183e+000?-7.71217883e-001
5.67980647e-001?1.74748570e-001?1.29376799e-001?-9.47380960e-001
-1.38257396e+000?1.08610772e-001?-2.31834769e+000?-3.18607867e-001
5.97291589e-001?2.97453463e-001?-2.88981467e-001?2.55147290e+000
4.35854554e-001?1.92910707e+000?-4.04442739e+000?3.35437202e+000
-7.18500257e-001?1.71847641e+000?3.76633972e-001?7.17979133e-001
-2.29244590e+000?4.33288908e+000?-3.21098715e-001?4.49396563e+000
-1.51703286e+000?-8.13243508e-001?3.68480325e-001?-2.04157472e+000
2.64199209e+000?-9.59263504e-001?-7.27866054e-001?-9.23123717e-001
1.79922330e+000?5.94993055e-001?1.60790789e+000?-1.55783963e+000
1.17599356e+000?3.75915971e-003?5.80935240e-001?-6.17859364e-001
5.47859287e+000?1.81236637e+000?-6.58016875e-002?1.29181659e+000
2.71043873e+000?1.00019448e-001?1.61559093e+000?1.83883631e+000
2.07489991e+000?8.18653643e-001?-7.93111801e-001?1.19132567e-002
4.03664291e-001?-1.44269729e+000?1.02415705e+000?2.56434369e+000
-7.33990967e-001?2.51126194e+000?-2.70398140e+000?-1.83867002e+000
1.18612683e+000?1.02612591e+000?-1.12655747e+000?-1.31735340e-001
-1.88770008e+000?-1.63585925e+000?-1.10935783e+000?1.27429748e+000
-5.52048720e-002?3.98856848e-001?-1.72617018e-001?-1.92361295e+000
-4.13408369e-001?-1.13245070e+000?2.85943294e+000?-2.07407784e+000
2.00722980e+000?7.59493232e-001?7.53989220e-001?5.31222701e-001
-1.06769693e+000?-7.28021516e-003?4.78438735e-001?-1.25011992e+000
-2.75211215e-001?-1.55662167e+000?-9.34187949e-001?-4.09402192e-001
1.53119981e+000?-1.25467312e+000?1.53636479e+000?-7.30930805e-001
2.75530076e+000?2.67432421e-001?2.85938501e+000?-1.42511964e+000
7.33835816e-001?1.99697030e+000?-3.78929794e-001?3.14493728e+000
1.20242715e+000?-2.51046824e+000?8.01128805e-001?-7.64064312e-001
1.19065619e+000?-7.97462463e-001?-4.37173009e-001?-3.08697790e-001
-2.21754000e-001?-2.32537538e-001?1.42117727e+000?1.26279199e+000
8.76116455e-001?-5.89745462e-001?-7.15154827e-001?-1.47839296e+000
3.58165920e-001?-1.45312059e+000?-1.79811522e-001?-1.22732484e+000
1.35544467e+000?-1.39437079e+000?4.28034633e-001?-4.00926471e-001
-4.24189746e-001?-1.87310791e+000?-7.61907935e-001?1.94510496e+000
-6.93394570e-003?-1.69082570e+000?1.55784622e-001?1.70695126e+000
1.48444736e+000?1.78316402e+000?6.74396753e-001?2.54230237e+000
3.12799782e-001?-2.27586889e+000?1.64418113e+000?-5.52159488e-001
1.09172642e+000?-1.33025026e+000?4.27963167e-001?-1.06191821e-002
2.02915812e+000?1.52300692e+000?1.75060666e+000?2.75519657e+000
9.04998899e-001?-1.99772191e+000?2.52548051e+000?-2.79550850e-001
2.24935460e+000?-3.23720884e+000?1.43948674e+000?-1.43738404e-001
2.14141667e-001?1.90374422e+000?-6.48318291e-001?-4.40555632e-001
1.67495656e+000?-1.21303253e-001?2.61465645e+000?3.95782739e-001
1.48943269e+000?-4.29212719e-001?2.21593046e+000?3.06588292e+000
-9.18957964e-002?1.64683378e+000?-2.01091081e-001?1.64259732e-001
1.44116330e+000?-1.27708030e+000?-5.94278574e-001?2.78909802e+000
4.04183298e-001?-6.39155686e-001?-1.11071789e+000?3.91723061e+000
8.88740361e-001?2.70166099e-001?1.68711150e+000?3.41494195e-002
2.62879491e+000?-1.22497030e-001?-8.07257146e-002?-1.23479038e-001
4.16377366e-001?1.49927616e-001?-9.35154483e-002?-2.42750645e-001
7.21661508e-001?6.92346573e-001?1.58245817e-001?-6.56240582e-001
2.31841183e+000?1.18762517e+000?-5.85088849e-001?1.60000634e+000
-4.65205479e+000?8.22968408e-002?-7.74762750e-001?-1.02885211e+000
1.38312590e+000?-8.51595581e-001?1.96487832e+000?2.25681591e+000
1.69004416e+000?3.35943174e+000?1.54889154e+000?3.73949504e+000
1.51178610e+000?-3.16800117e-001?2.04924011e+000?8.30704570e-001
1.65477479e+000?-1.30295336e+000?4.55465412e+000?-1.38525236e+000
1.86339700e+000?1.91202164e+000?9.32169855e-001?-5.06053150e-001
-4.11295146e-002?6.56290650e-001?1.72126293e+000?-2.35860395e+000
1.45457065e+000?-3.16340141e-002?-2.53601027e+000?1.32477188e+000
7.35908270e-001?-3.98547322e-001?1.66871738e+000?3.61932486e-001
-1.42626956e-001?2.08667588e+000?1.06165230e+000?-1.04248536e+000
1.65681597e-002?1.52931511e+000?2.51952916e-001?-1.65126300e+000
1.22786045e+000?-2.75025427e-001?6.26901507e-001?-1.21067852e-001
1.91679001e+000?-1.16014004e+000?3.37646812e-001?3.64376120e-002
-2.23688388e+000?-1.95473611e+000?1.15201533e+000?-3.13668609e+000
6.84875786e-001?-1.19674683e+000?-1.24230969e+000?-2.26241636e+000
-5.24263144e-001?3.58203673e+000?2.40704998e-001?9.85608280e-001
7.78039932e-001?-7.78025836e-002?5.06118417e-001?-1.16267717e+000
-2.90982664e-001?1.29055846e+000?-4.62567091e-001?3.04141134e-001
3.29742098e+000?-1.57454121e+000?-6.33357108e-001?1.04142857e+000
-8.39529216e-001?1.97137713e+000?1.80137825e+000?-5.80883324e-001
8.53551626e-001?-2.44389606e+000?-1.04786527e+000?6.01655066e-001
1.35108554e+000?1.91133451e+000?-1.55116379e+000?-1.32689738e+000
-3.34982187e-001?-6.04158759e-001?-7.12676525e-001?1.43757379e+000
-3.67825031e-001?1.35954702e+000?5.94304781e-003?-1.79859340e+000
-2.72952080e+000?8.32889140e-001?-1.39144945e+000?-1.17416644e+000
-7.64998674e-001?4.93923604e-001?-1.44545782e+000?1.57880306e+000
1.12437308e+000?1.19925821e+000?7.10867286e-001?4.67830688e-001
6.74548447e-001?1.09623861e+000?2.23827716e-002?-3.27283883e+000
-1.48206964e-001?4.71823335e-001?9.59278345e-001?2.86045492e-001
9.12338674e-001?-2.33875215e-001?1.37641966e-001?2.69536877e+000
2.83514071e+000?-2.39875865e+000?4.55236256e-001?2.48535085e+000
2.72361922e+000?2.54125237e-001?-3.18247676e+000?1.55454111e+000
2.03626370e+000?-6.98356092e-001?-3.20186406e-001?3.07346135e-001
5.24029970e-001?3.58576584e+000?-8.57803285e-001?1.20365873e-001
5.76087952e-001?4.73395920e+000?-4.21760499e-001?2.81904149e+000
1.39611915e-001?-2.31253937e-001?-1.23439991e+000?1.50822794e+000
9.40651536e-001?-7.68359900e-001?2.31346178e+000?-5.87843180e-001
1.24623096e+000?-2.88376665e+000?9.03283894e-001?-1.24732837e-001
2.60412898e-002?2.28092027e+000?-7.80223548e-001?1.42146301e+000
1.28785765e+000?-7.78531671e-001?-1.10662627e+000?3.81465006e+000
1.45004451e+000?-6.86295390e-001?1.31134605e+000?-7.53161073e-001
-2.38736463e+000?9.48513985e-001?2.12668252e+000?2.01590508e-001
2.64392781e+000?-2.63787150e-001?-2.52131611e-001?5.42520463e-001
2.20529389e+000?2.12599683e+000?4.67926890e-001?-1.01627827e+000
-4.38636959e-001?-1.01159893e-001?-6.70790598e-002?1.14079630e+000
2.45170498e+000?-9.42526609e-002?1.41221488e+000?2.27625277e-002
-2.80443698e-001?1.27343559e+000?2.37848088e-001?1.82034159e+000
-8.26801598e-001?-7.50176728e-001?-1.08648551e+000?2.42070699e+000
2.27270246e+000?-1.19815774e-001?-9.50998545e-001?1.68236101e+000
-2.97976494e+000?9.55732107e-001?2.90135597e-003?-2.69438386e+000
3.41107798e+000?1.11490011e+000?-5.75072527e-001?3.25586051e-001
1.13065195e+000?9.45354939e-001?-3.86584491e-001?-1.10546327e+000
1.85765481e+000?-3.66105199e+000?-1.34011900e+000?-3.60621244e-001
-1.27096891e+000?-7.11437106e-001?2.39505959e+000?1.65358496e+000
1.50321412e+000?3.16486776e-001?4.79341358e-001?1.55303967e+000
3.40338850e+000?4.89003092e-001?2.81465554e+000?-5.24295390e-001
-1.56343615e+000?-1.18851018e+000?3.45909023e+000?5.30569851e-001
1.25108421e+000?-3.48280728e-001?-2.72895396e-001?1.10347736e+000
7.20215678e-001?-2.27794456e+000?-1.01811433e+000?-3.29465437e+000
1.60887623e+000?4.71850276e-001?-2.30937099e+000?-1.09952450e+000
-2.89444745e-001?1.14027429e+000?3.20806056e-001?4.92361672e-002
9.30287540e-001?1.52699798e-001?-1.64544094e+000?1.50056016e+000
-1.19767833e+000?2.22963786e+000?2.25639606e+000?9.56510782e-001
1.05498433e+000?2.60551929e+000?2.47192788e+000?-1.94899297e+000
9.73172724e-001?6.81347847e-001?3.84335667e-002?5.63789785e-001
2.09292388e+000?-2.35492945e+000?-1.60491121e+000?-4.33678776e-001
2.58014894e+000?3.53837281e-001?1.20613635e+000?-1.03824782e+000
-2.70244169e+000?4.62901741e-001?-5.16113460e-001?-2.26867557e+000
7.69846320e-001?-1.06869304e+000?-2.49847388e+000?1.66551188e-001
-2.11130619e-001?1.27242219e+000?2.40089417e+000?-1.50180638e+000
2.83866119e+000?1.66601646e+000?1.21163785e+000?1.51577842e+000
3.16205978e-001?-4.14970517e-001?3.34209830e-001?1.90904868e+000
-6.08527839e-001?2.00294709e+000?-1.42225236e-001?2.44074321e+000
-2.83460450e+000?3.70286083e+000?1.55271268e+000?-1.60934865e-001
-2.05457234e+000?-1.11318493e+000?7.83119619e-001?1.93219930e-001
6.09611750e-001?-8.89916360e-001?-5.18336222e-002?1.62632263e+000
2.05337271e-001?2.36734167e-001?1.08341157e+000?-8.27812910e-001
-1.50358784e+000?7.58716226e-001?-2.18823433e+000?-3.84354234e+000
-7.46470034e-001?1.35967243e+000?-1.51201522e+000?6.61429822e-001
4.33022141e-001?1.31914115e+000?-2.84156471e-001?3.01826668e+000
8.90323818e-002?2.57224154e+000?-3.49536806e-001?1.20527959e+000
-1.64551783e+000?5.97079456e-001?1.46535718e+000?-7.79165745e-001
2.56506968e+000?1.64554584e+000?2.43936133e+000?-1.24797320e+000
1.61060110e-001?2.62977779e-001?6.60554469e-001?-1.77881193e+000
-1.68319678e+000?-3.18312883e-001?-6.25143230e-001?-1.05535913e+000
4.94530886e-001?-5.20972371e-001?-2.16873646e+000?-1.80824995e+000
3.11055136e+000?-3.26495266e+000?-2.33749151e+000?8.06113109e-002
-7.43417799e-001?-1.43765950e+000?1.28656340e+000?9.37331021e-001
1.65003514e+000?2.17272830e+000?5.64124525e-001?3.01361680e+000
1.67168570e+000?2.28130484e+000?1.23492706e+000?6.32456541e-001
1.87143099e+000?-1.76699713e-001?6.86713874e-001?-6.65625334e-002
2.96964973e-001?3.82044941e-001?-3.73141348e-001?1.65588665e+000
-1.86311111e-001?-3.70510578e+000?2.68024445e+000?-2.11436319e+000
2.71787930e+000?-7.34579444e-001?-2.32560158e+000?1.64193237e+000
1.98097467e+000?1.09497666e+000?5.11818349e-001?-3.73851085e+000
4.02708441e-001?2.30055690e+000?-3.67630422e-001?-3.54356766e+000
9.37409978e-003?-1.08068323e+000?2.79761529e+000?6.36268973e-001
-1.53954104e-001?4.92218137e-001?1.00797129e+000?-1.79180205e+000
2.27602243e+000?2.52540797e-001?1.12398112e+000?-2.50184131e+000
1.30374336e+000?1.00840843e+000?7.91947007e-001?1.65224433e-001
-7.91436791e-001?2.79811144e+000?3.53997409e-001?-3.11636615e+000
-2.11673722e-001?4.85001028e-001?2.78055859e+000?-1.47256184e+000
2.09027216e-001?-4.01711315e-001?-2.80359674e+000?7.92990699e-002
-3.77475917e-001?3.30463171e-001?5.17215729e-001?-2.90217614e+000
7.09646821e-001?1.22404337e+000?1.48780787e+000?3.30003649e-001
7.30821073e-001?-3.11456323e+000?4.59069252e-001?-2.16302824e+000
1.05775273e+000?6.04678869e-001?-1.66926908e+000?6.87627673e-001
-1.33886015e+000?-2.73541957e-001?-7.91138187e-002?-4.61171776e-001
-2.02059436e+000?-1.02667737e+000?1.42993629e+000?1.25154507e+000
-5.92566133e-001?1.91667759e+000?1.59074605e+000?1.06317270e+000
3.81611019e-001?-9.54510391e-001?3.52517843e+000?-4.50870705e+000
1.48635769e+000?-2.68166208e+000?-8.87980342e-001?1.16167498e+000
1.40314496e+000?4.81270671e-001?1.04817164e+000?-1.26742125e+000
3.31832290e+000?-2.09496140e+000?-8.00072968e-001?7.72442520e-001
-2.10208988e+000?-8.34703386e-001?-3.58576655e-001?1.82497489e+000
-1.93366349e+000?1.48226321e+000?-7.61157155e-001?4.60760474e-001
-2.66458559e+000?7.54581615e-002?-1.02266443e+000?-3.88942146e+000
-2.46707106e+000?1.35353601e+000?-3.33155775e+000?-5.69405317e-001
-2.34355664e+000?9.14932430e-001?2.90386176e+000?-9.07698750e-001
2.30322170e+000?-5.37009060e-001?3.27328682e+000?-2.76939601e-001
9.18692470e-001?1.32038641e+000?3.33021188e+000?1.15975368e+000
1.04361510e+000?-2.81390834e+000?-2.22346139e+000?-1.06822634e+000
6.44918501e-001?-4.19576801e-002?-9.63480592e-001?-1.70784652e+000
-4.60445702e-001?-1.37904525e+000?3.25156331e-001?5.62403738e-001
1.78129339e+000?1.33165371e+000?-1.49321783e+000?-1.57136106e+000
9.94577527e-001?4.57476169e-001?2.44409823e+000?-5.55334091e-001
2.25377584e+000?1.19198716e+000?-2.42960095e+000?-1.06827486e+000
3.08647299e+000?1.74202055e-001?-2.06566954e+000?-4.41996425e-001
8.30268204e-001?-3.33665919e+000?-9.09507215e-001?2.08288980e+000
1.35339153e+000?-7.76457965e-001?-8.69336069e-001?2.39780879e+000
2.45157194e+000?-3.29353005e-001?1.56075108e+000?1.19201887e+000
1.26750851e+000?8.99221838e-001?1.90738845e+000?1.81938136e+000
-9.63269055e-001?-7.57165074e-001?1.55659997e+000?-1.92750478e+000
-1.79447556e+000?7.61612654e-001?-1.06798351e+000?-1.52491510e+000
8.09498847e-001?8.15360665e-001?1.45587683e+000?-2.96532065e-001
3.49902302e-001?1.52817357e+000?-3.48740928e-002?8.55526030e-001
-4.30970520e-001?-1.73958808e-001?7.96336412e-001?-6.07606351e-001
-4.60522503e-001?9.14190173e-001?5.89256465e-001?-1.46241927e+000
1.27102017e+000?7.72446930e-001?8.97496998e-001?-2.73194551e-001
-2.78902292e+000?3.15747571e+000?1.36864412e+000?-1.71781838e+000
1.42213500e+000?7.46331871e-001?-9.78936791e-001?-3.39532375e+000
1.66855657e+000?1.07649493e+000?-9.87916648e-001?-9.83610451e-001
1.10034108e+000?4.18890901e-002?3.20622015e+000?3.36752504e-001
2.29340529e+000?-1.53924727e+000?5.37710667e-001?-1.16541696e+000
6.92831993e-001?-1.17176509e+000?2.06588030e+000?1.45995200e+000
-1.28829181e+000?1.19395006e+000?-1.74698621e-001?2.01432729e+000
3.77614737e-001?-1.18296631e-001?-7.40564585e-001?9.01206136e-001
-7.46783733e-001?5.06496355e-002?-1.10556877e+000?2.90351462e+000
1.33163059e+000?-6.31340027e-001?-1.03813303e+000?-1.19116557e+000
-5.86888194e-001?1.15237069e+000?1.42651880e+000?-1.35818005e+000
-2.57812238e+000?3.82667094e-001?6.12372935e-001?1.48428476e+000
1.35794133e-001?7.76758254e-001?-7.03539670e-001?-6.03174090e-001
-6.03082001e-001?1.99687672e+000?-1.18074071e+000?-1.21837628e+000
-2.08009586e-001?-1.05319881e+000?2.63637996e+000?6.30942523e-001
5.15693784e-001?1.76138854e+000?-7.68377483e-001?-4.59355921e-001
6.46186352e-001?-1.86447725e-001?1.31977904e+000?-1.21378517e+000
8.48503828e-001?-1.69902217e+000?6.27434492e-001?-1.20918322e+000
-2.71898794e+000?6.21949911e-001?1.20734927e-004?-2.41284895e+000
-1.65921319e+000?7.24378884e-001?-7.34631896e-001?-2.36548260e-001
4.22343224e-001?1.13314591e-001?-1.44387603e+000?3.21988726e+000
9.78636324e-001?-2.25895897e-001?1.36732399e+000?-1.27787903e-001
-1.14109528e+000?4.24284172e+000?3.68867755e-001?-3.94058943e-001
-1.87558246e+000?-3.67672849e+000?1.35180593e+000?5.44494271e-001
-2.11900043e+000?-8.52134079e-002?-1.05214870e+000?-1.32194981e-001
-1.45356727e+000?3.20472717e-002?6.06840193e-001?-1.63723218e+000
9.62877572e-001?-4.61500555e-001?9.63598549e-001?-8.61029863e-001
-1.06589496e+000?-4.52161729e-001?-8.19720447e-001?1.20012617e+000
1.91720557e+000?2.12556529e+000?-1.88315898e-001?-6.26728415e-001
-6.68102264e-001?1.10652697e+000?2.58852243e-001?2.66154432e+000
-7.34255850e-001?-3.12065445e-002?-2.22151279e+000?-1.37975235e-002
-1.51210678e+000?-9.28165853e-001?4.08833265e+000?-1.14312494e+000
1.29504907e+000?2.74469674e-001?1.71046054e+000?6.18974745e-001
-2.06761813e+000?6.44383669e-001?1.23289537e+000?1.67420253e-001
1.30074763e+000?-1.73213923e+000?4.22435611e-001?-6.98242247e-001
3.73503900e+000?-9.40370619e-001?4.61774921e+000?-1.93580663e+000
-5.14993525e+000?1.67973175e+001?-2.17262897e+001?-9.32018375e+000
-1.15877132e+001?1.62843156e+000
Suppose that the HOG feature x calculating a fingerprint image is as follows, it is one 1296 dimension data:
The HOG feature x of this fingerprint image is input to optimal classification surface function, and being namely updated to following formula can obtain: f (x)=wx-b=2.1761881 > 0.Therefore can judge that the HOG feature of this fingerprint image drops on f (x) > 0 side, so can judge that it is the measured fingerprint image of matter, output label value y=+1.
Suppose that the HOG feature x calculating another fingerprint image is as follows, it is one 1296 dimension data:
The HOG feature x of this fingerprint image is input to optimal classification surface function, and being namely updated to following formula can obtain: f (x)=wx-b=-1.7858665 < 0.Therefore can judge that the HOG feature of this fingerprint image drops on f (x) < 0 side, so can judge that it is ropy fingerprint image, output label value y=-1.
Thus, the determination methods of fingerprint image quality of the present invention, by HOG feature being incorporated in fingerprint image quality judgement, adopting the mode determination optimal classification surface of machine learning, judging the quality of fingerprint image quality according to the HOG characteristic sum optimal classification surface of fingerprint image.Not only save the work manually determining judgment threshold, and there is good extended capability, namely this judgment mode can judge the impact that polytype noise brings, and the sample type that only need input needs can complete judgement, shows splendid effect in great many of experiments.Determination methods of the present invention is mainly used in small size fingerprint collecting equipment and judges the fingerprint image quality quality gathered, especially for the situation that sweat stain, mud stain, noise etc. cause fingerprint image fuzzy, can judge accurately the quality of fingerprint image before fingerprint recognition, for improving fingerprint recognition rate, reduction accuracy of system identification is laid a good foundation.
See Fig. 5, Fig. 6, propose judgment means one embodiment of fingerprint image quality of the present invention, described judgment means comprises study module and judge module.
Study module: for obtaining fingerprint image sample, utilizes SVM classifier to learn according to fingerprint image sample, obtains optimal classification surface, and the function of optimal classification surface and correlation parameter are sent to judge module.
Fingerprint image sample comprises positive and negative sample, i.e. the measured sample of matter and ropy sample, at least each one of positive negative sample, preferably multiple.Fingerprint image sample is by artificial selection, and can be the on-the-spot artificial fingerprint image gathering acquisition, also can be the ready-made fingerprint image obtained from outside.
The HOG feature of study module first calculated fingerprint image pattern, then carries out training study by the HOG feature of fingerprint image sample input SVM classifier, thus obtains optimal classification surface.
When calculating HOG feature, study module is the gradient direction value of each location of pixels in calculated fingerprint image pattern first; Then fingerprint image sample being divided into multiple cell, is that each cell builds gradient orientation histogram according to gradient direction value; Then cell is combined into block, normalized gradient direction histogram in block; Finally the gradient orientation histogram of all pieces in fingerprint image sample is combined and form HOG feature.
The input of the HOG feature of multiple fingerprint image sample SVM, SVM are then carried out training study according to the HOG feature of multiple fingerprint image sample by study module, obtain optimal classification surface.The function of this optimal classification surface is f (x)=wx-b, and wherein w is the support vector of optimal classification surface, and b is constant term, and x is the HOG feature of the image of input, w and b is the parameter of optimal classification surface.
Judge module: for obtaining fingerprint image to be judged, and calculate the HOG feature of this fingerprint image, judge the quality of fingerprint image according to HOG characteristic sum optimal classification surface, and export judged result.
Judge module comprises processing unit and judgement unit, wherein:
Processing unit: for obtaining fingerprint image to be judged, the HOG feature of calculated fingerprint image, calculates in the function of this HOG feature input optimal classification surface, and sends judged result to judgement unit.
Concrete, processing unit is the gradient direction value of each location of pixels in calculated fingerprint image first; Then fingerprint image being divided into multiple cell, is that each cell builds gradient orientation histogram according to gradient direction value; Then cell is combined into block, normalized gradient direction histogram in block, the gradient orientation histogram of all pieces in fingerprint image is combined and forms HOG feature.Finally, processing unit calculates in function f (the x)=w*x-b of the HOG feature of fingerprint image input optimal classification surface, and wherein w is the support vector of optimal classification surface, and b is constant term, and x is the HOG feature of fingerprint image.
Judgement unit: differentiate for the result of calculation according to processing unit, if result of calculation is f (x) > 0, then judges that the quality of fingerprint image is good, output label value y=+1; If result of calculation is f (x) < 0, then judge the of poor quality of fingerprint image, output label value y=-1.
Accordingly, the judgment means of fingerprint image quality of the present invention, by introducing HOG feature and adopting SVM supervised learning pattern successfully to distinguish fingerprint image quality, especially for the situation that sweat stain, mud stain, noise etc. cause fingerprint image fuzzy, can judge accurately the quality of fingerprint image before fingerprint recognition, for improving discrimination, reduction accuracy of system identification is laid a good foundation, and is particularly useful for small size fingerprint collecting equipment.Judgment means of the present invention, not only save the work manually determining judgment threshold, and there is good extended capability, namely this judgment mode can judge the impact that polytype noise brings, the sample type that only need input needs can complete judgement, shows splendid effect in great many of experiments.
It should be noted that: the judgment means of the fingerprint image quality that above-described embodiment provides is when carrying out the judgement of fingerprint image quality, only be illustrated with the division of above-mentioned each functional module, in practical application, can distribute as required and by above-mentioned functions and be completed by different functional modules.In addition, the judgment means of the fingerprint image quality that above-described embodiment provides and the determination methods embodiment of fingerprint image quality belong to same design, its specific implementation process refers to embodiment of the method, and the technical characteristic in embodiment of the method is all corresponding applicable in device embodiment, repeats no more here.
One of ordinary skill in the art will appreciate that, realize the hardware that all or part of step in above-described embodiment method can control to be correlated with by program to complete, described program can be stored in a computer read/write memory medium, and described storage medium can be ROM/RAM, disk, CD etc.
Should be understood that; these are only the preferred embodiments of the present invention; can not therefore limit the scope of the claims of the present invention; every utilize instructions of the present invention and accompanying drawing content to do equivalent structure or equivalent flow process conversion; or be directly or indirectly used in other relevant technical fields, be all in like manner included in scope of patent protection of the present invention.
Claims (8)
1. a determination methods for fingerprint image quality, is characterized in that, comprises step:
Obtain fingerprint image sample;
Utilize SVM support vector machine classifier to learn according to described fingerprint image sample, obtain optimal classification surface;
Obtain fingerprint image to be judged, and calculate the HOG gradient orientation histogram feature of described fingerprint image;
The quality of described fingerprint image is judged according to described HOG characteristic sum optimal classification surface.
2. the determination methods of fingerprint image quality according to claim 1, is characterized in that, the described SVM of utilization support vector machine classifier carries out study acquisition optimal classification surface according to described fingerprint image sample and comprises:
Calculate the HOG feature of described fingerprint image sample;
The HOG feature of described fingerprint image sample is inputted in described SVM classifier and carries out training study, obtain optimal classification surface.
3. the determination methods of fingerprint image quality according to claim 1, is characterized in that, the HOG gradient orientation histogram feature of the described fingerprint image of described calculating comprises:
Calculate the gradient direction value of each location of pixels in described fingerprint image;
Described fingerprint image is divided into multiple cell, is that each cell builds gradient orientation histogram according to described gradient direction value;
Described cell is combined into block, gradient orientation histogram described in normalization in described piece, the gradient orientation histogram of all pieces in described fingerprint image is combined and forms HOG feature.
4. the determination methods of fingerprint image quality according to claim 1, is characterized in that, describedly judges that the quality of described fingerprint image comprises according to described HOG characteristic sum optimal classification surface:
The HOG feature of described fingerprint image inputted in function f (the x)=w*x-b of described optimal classification surface and calculate, wherein w is the support vector of optimal classification surface, and b is constant term, and x is the HOG feature of fingerprint image;
If result of calculation is f (x) > 0, then judge that the quality of described fingerprint image is good;
If result of calculation is f (x) < 0, then judge the of poor quality of described fingerprint image.
5. a judgment means for fingerprint image quality, is characterized in that, comprises study module and judge module, wherein:
Study module, for obtaining fingerprint image sample, utilizes SVM classifier to learn according to described fingerprint image sample, obtains optimal classification surface;
Judge module, for obtaining fingerprint image to be judged, and calculating the HOG gradient orientation histogram feature of described fingerprint image, judging the quality of described fingerprint image according to described HOG characteristic sum optimal classification surface.
6. the judgment means of fingerprint image quality according to claim 5, it is characterized in that, described study module is used for: the HOG feature calculating described fingerprint image sample, the HOG feature of described fingerprint image sample is inputted in described SVM classifier and carries out training study, obtains optimal classification surface.
7. the judgment means of fingerprint image quality according to claim 5, is characterized in that, described judge module comprises processing unit, and described processing unit is used for: the gradient direction value calculating each location of pixels in described fingerprint image; Described fingerprint image is divided into multiple cell, is that each cell builds gradient orientation histogram according to described gradient direction value; Described cell is combined into block, gradient orientation histogram described in normalization in described piece, the gradient orientation histogram of all pieces in described fingerprint image is combined and forms HOG feature.
8. the judgment means of fingerprint image quality according to claim 5, is characterized in that, described judge module comprises processing unit and judgement unit, wherein:
Processing unit, for the HOG feature of described fingerprint image is inputted described optimal classification surface function f (x)=w*x-b in calculate, wherein w is the support vector of optimal classification surface, and b is constant term, and x is the HOG feature of fingerprint image;
Judgement unit, differentiates for the result of calculation according to processing unit, if result of calculation is f (x) > 0, then judges that the quality of described fingerprint image is good; If result of calculation is f (x) < 0, then judge the of poor quality of described fingerprint image.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410510227.0A CN104268529A (en) | 2014-09-28 | 2014-09-28 | Judgment method and device for quality of fingerprint images |
PCT/CN2014/094567 WO2016045215A1 (en) | 2014-09-28 | 2014-12-22 | Fingerprint image quality determination method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410510227.0A CN104268529A (en) | 2014-09-28 | 2014-09-28 | Judgment method and device for quality of fingerprint images |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104268529A true CN104268529A (en) | 2015-01-07 |
Family
ID=52160049
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410510227.0A Pending CN104268529A (en) | 2014-09-28 | 2014-09-28 | Judgment method and device for quality of fingerprint images |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN104268529A (en) |
WO (1) | WO2016045215A1 (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105528591A (en) * | 2016-01-14 | 2016-04-27 | 电子科技大学 | Living fingerprint identification method based on multi-quadrant coding |
CN105718848A (en) * | 2015-10-21 | 2016-06-29 | 深圳芯启航科技有限公司 | Quality evaluation method and apparatus of fingerprint images |
CN106355145A (en) * | 2016-08-26 | 2017-01-25 | 广东欧珀移动通信有限公司 | Fingerprint identification method, device and mobile terminal |
CN106408552A (en) * | 2015-07-27 | 2017-02-15 | 辽宁省公安厅 | Multi-finger position fingerprint image quality comprehensive evaluation method based on rank sum ratio method |
CN106529545A (en) * | 2016-09-26 | 2017-03-22 | 北京林业大学 | Speckle image quality recognition method and system based on image feature description |
CN108985351A (en) * | 2018-06-27 | 2018-12-11 | 北京中安未来科技有限公司 | It is a kind of that the method and apparatus of blurred picture are identified based on gradient direction sparse features information, calculate equipment and storage medium |
CN110472518A (en) * | 2019-07-24 | 2019-11-19 | 杭州晟元数据安全技术股份有限公司 | A kind of fingerprint image quality judgment method based on full convolutional network |
CN111428748A (en) * | 2020-02-20 | 2020-07-17 | 重庆大学 | Infrared image insulator recognition and detection method based on HOG characteristics and SVM |
CN112699863A (en) * | 2021-03-25 | 2021-04-23 | 深圳阜时科技有限公司 | Fingerprint enhancement algorithm, computer-readable storage medium and electronic device |
CN115841685A (en) * | 2023-02-15 | 2023-03-24 | 南京信息工程大学 | Fake fingerprint detection system and method based on composite pixel gradient |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110838116B (en) * | 2019-11-14 | 2023-01-03 | 上海联影医疗科技股份有限公司 | Medical image acquisition method, device, equipment and computer-readable storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101303728A (en) * | 2008-07-01 | 2008-11-12 | 山东大学 | Method for identifying fingerprint facing image quality |
CN102831409A (en) * | 2012-08-30 | 2012-12-19 | 苏州大学 | Method and system for automatically tracking moving pedestrian video based on particle filtering |
US20140226024A1 (en) * | 2013-02-08 | 2014-08-14 | Kutta Technologies, Inc. | Camera control in presence of latency |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7853072B2 (en) * | 2006-07-20 | 2010-12-14 | Sarnoff Corporation | System and method for detecting still objects in images |
CN102169533A (en) * | 2011-05-11 | 2011-08-31 | 华南理工大学 | Commercial webpage malicious tampering detection method |
CN102799669B (en) * | 2012-07-17 | 2015-06-17 | 杭州淘淘搜科技有限公司 | Automatic grading method for commodity image vision quality |
CN102915446A (en) * | 2012-09-20 | 2013-02-06 | 复旦大学 | Plant disease and pest detection method based on SVM (support vector machine) learning |
CN102982350B (en) * | 2012-11-13 | 2015-10-28 | 上海交通大学 | A kind of station caption detection method based on color and histogram of gradients |
-
2014
- 2014-09-28 CN CN201410510227.0A patent/CN104268529A/en active Pending
- 2014-12-22 WO PCT/CN2014/094567 patent/WO2016045215A1/en active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101303728A (en) * | 2008-07-01 | 2008-11-12 | 山东大学 | Method for identifying fingerprint facing image quality |
CN102831409A (en) * | 2012-08-30 | 2012-12-19 | 苏州大学 | Method and system for automatically tracking moving pedestrian video based on particle filtering |
US20140226024A1 (en) * | 2013-02-08 | 2014-08-14 | Kutta Technologies, Inc. | Camera control in presence of latency |
Non-Patent Citations (1)
Title |
---|
慕春雷: "基于HOG特征的人脸识别系统研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106408552A (en) * | 2015-07-27 | 2017-02-15 | 辽宁省公安厅 | Multi-finger position fingerprint image quality comprehensive evaluation method based on rank sum ratio method |
CN106408552B (en) * | 2015-07-27 | 2019-07-12 | 辽宁省公安厅 | A kind of more finger position fingerprint image quality integrated evaluating methods based on sum of ranks ratio method |
CN105718848B (en) * | 2015-10-21 | 2020-08-07 | 深圳芯启航科技有限公司 | Quality evaluation method and device for fingerprint image |
CN105718848A (en) * | 2015-10-21 | 2016-06-29 | 深圳芯启航科技有限公司 | Quality evaluation method and apparatus of fingerprint images |
CN105528591A (en) * | 2016-01-14 | 2016-04-27 | 电子科技大学 | Living fingerprint identification method based on multi-quadrant coding |
CN105528591B (en) * | 2016-01-14 | 2019-04-16 | 电子科技大学 | Living body finger print recognition methods based on more quadrants coding |
CN106355145A (en) * | 2016-08-26 | 2017-01-25 | 广东欧珀移动通信有限公司 | Fingerprint identification method, device and mobile terminal |
CN106529545A (en) * | 2016-09-26 | 2017-03-22 | 北京林业大学 | Speckle image quality recognition method and system based on image feature description |
CN108985351B (en) * | 2018-06-27 | 2021-11-26 | 北京中安未来科技有限公司 | Method and device for recognizing blurred image based on gradient direction sparse characteristic information, computing equipment and storage medium |
CN108985351A (en) * | 2018-06-27 | 2018-12-11 | 北京中安未来科技有限公司 | It is a kind of that the method and apparatus of blurred picture are identified based on gradient direction sparse features information, calculate equipment and storage medium |
CN110472518A (en) * | 2019-07-24 | 2019-11-19 | 杭州晟元数据安全技术股份有限公司 | A kind of fingerprint image quality judgment method based on full convolutional network |
CN110472518B (en) * | 2019-07-24 | 2022-05-17 | 杭州晟元数据安全技术股份有限公司 | Fingerprint image quality judgment method based on full convolution network |
CN111428748A (en) * | 2020-02-20 | 2020-07-17 | 重庆大学 | Infrared image insulator recognition and detection method based on HOG characteristics and SVM |
CN112699863A (en) * | 2021-03-25 | 2021-04-23 | 深圳阜时科技有限公司 | Fingerprint enhancement algorithm, computer-readable storage medium and electronic device |
CN112699863B (en) * | 2021-03-25 | 2022-05-17 | 深圳阜时科技有限公司 | Fingerprint enhancement method, computer-readable storage medium and electronic device |
CN115841685A (en) * | 2023-02-15 | 2023-03-24 | 南京信息工程大学 | Fake fingerprint detection system and method based on composite pixel gradient |
Also Published As
Publication number | Publication date |
---|---|
WO2016045215A1 (en) | 2016-03-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104268529A (en) | Judgment method and device for quality of fingerprint images | |
CN108319964B (en) | Fire image recognition method based on mixed features and manifold learning | |
CN104424466B (en) | Method for checking object, body detection device and image pick up equipment | |
CN103077407B (en) | Car logo positioning and recognition method and car logo positioning and recognition system | |
Yao et al. | Traffic sign recognition using HOG-SVM and grid search | |
CN104063722A (en) | Safety helmet identification method integrating HOG human body target detection and SVM classifier | |
CN105426905A (en) | Robot barrier identification method based on gradient histogram and support vector machine | |
CN113435546B (en) | Migratable image recognition method and system based on differentiation confidence level | |
CN102567994B (en) | Infrared small target detection method based on angular point gaussian characteristic analysis | |
CN104268528A (en) | Method and device for detecting crowd gathered region | |
CN105279772A (en) | Trackability distinguishing method of infrared sequence image | |
CN102945374B (en) | Method for automatically detecting civil aircraft in high-resolution remote sensing image | |
CN103679186A (en) | Target detecting and tracking method and device | |
CN105069774A (en) | Object segmentation method based on multiple-instance learning and graph cuts optimization | |
CN107886066A (en) | A kind of pedestrian detection method based on improvement HOG SSLBP | |
CN106485273A (en) | A kind of method for detecting human face based on HOG feature and DNN grader | |
CN106780727A (en) | A kind of headstock detection model method for reconstructing and device | |
CN103310235A (en) | Steganalysis method based on parameter identification and estimation | |
CN107886539A (en) | High class gear visible detection method under a kind of industrial scene | |
CN104123714A (en) | Optimal target detection scale generation method in people flow statistics | |
CN115439458A (en) | Industrial image defect target detection algorithm based on depth map attention | |
CN115131747A (en) | Knowledge distillation-based power transmission channel engineering vehicle target detection method and system | |
CN105809092A (en) | Population target detection method and device thereof | |
CN104331708B (en) | A kind of zebra crossing automatic detection analysis method and system | |
CN116579616B (en) | Risk identification method based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150107 |