CN106529407A - Vehicle-mounted fingerprint recognition method - Google Patents
Vehicle-mounted fingerprint recognition method Download PDFInfo
- Publication number
- CN106529407A CN106529407A CN201610880701.8A CN201610880701A CN106529407A CN 106529407 A CN106529407 A CN 106529407A CN 201610880701 A CN201610880701 A CN 201610880701A CN 106529407 A CN106529407 A CN 106529407A
- Authority
- CN
- China
- Prior art keywords
- fingerprint
- image
- point
- digital picture
- vehicle
- 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
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
- G06V40/13—Sensors therefor
-
- 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
- G06V40/1347—Preprocessing; Feature extraction
-
- 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
Abstract
The invention belongs to the technical field of identity verification for vehicle theft prevention, and specifically relates to a vehicle-mounted fingerprint recognition method. The method comprises the steps: fingerprint collection, fingerprint digital image preprocessing, fingerprint feature extraction, and fingerprint feature matching. The fingerprint digital image preprocessing comprises the steps: imaging sharpening; gray scale transformation; image binaryzation; image thinning. The fingerprint feature extraction comprises the steps: end point identification; bifurcated point identification; and removal of a false result. The method improves the precision of fingerprint recognition.
Description
Technical field
The invention belongs to the identity validation technology field of automobile burglar, specifically, is related to a kind of vehicle-mounted fingerprint identification method.
Background technology
The basic role one of automobile burglar product is to take precautions against vehicle to be illegally used without permission, current vapour vehicle anti-theft system
Unite predominantly organic tool anti-theft device, burglar alarm, networking guarding against theft warning system and based on fingerprint recognition prevent
Theft system.
Fingerprint recognition system includes fingerprint collecting, fingerprint digital picture pretreatment, Finger print characteristic abstract and fingerprint characteristic
It is most of with four.
At present, the precision of fingerprint digital picture identification is not very high, and reason is exactly that anti-theft system installed in car is calculated to fingerprint recognition
The performance requirement of method is high, Image semantic classification is not carried out, and Finger print characteristic abstract is inaccurately caused.
The content of the invention
It is an object of the invention to provide a kind of vehicle-mounted fingerprint identification method, to solve the above problems.
The embodiment provides a kind of vehicle-mounted fingerprint identification method, including：Fingerprint collecting, fingerprint digital picture are pre-
Process, Finger print characteristic abstract and fingerprint minutiae matching；Wherein, fingerprint digital picture pretreatment includes：Image sharpening；Gray scale becomes
Change；Image binaryzation；Image thinning；Finger print characteristic abstract includes Endpoint ID；Bifurcated point identification；Remove dummy results.
Further, fingerprint digital picture pretreatment also includes：Fingerprint image gray scale normalization；Fingerprint image gray balance
Change；Fingerprint Image Segmentation；Gabor is filtered.
Further, image sharpening includes：Wave filter is generated using fspecial functions, imfilter function pair figures are recycled
As carrying out convolutional calculation；Wherein fspecial type function is unsharp types.
Further, image thinning includes：Image is changed into into single pixel connected graph by the way of template matching.
Further, removing dummy results includes：The point of image border is cut off using the method that fingerprint image cuts；Using away from
Characteristic point closer to the distance is removed from threshold method.
Further, Fingerprint Image Segmentation is using the fingerprint segmentation method based on gradient.
Further, fingerprint minutiae matching includes：Extract the intermediate point of fingerprint digital picture to be matched, using the intermediate point as
Matching reference minutiae；Using end points and the characteristic vector template of intersection feature construction warehouse-in fingerprint；By matching reference minutiae and template
Match point in storehouse is contrasted, and weighs fingerprint similarity using Euclidean distance, calculates whether fingerprint matches.
Compared with prior art the invention has the beneficial effects as follows：Improve the degree of accuracy of fingerprint recognition.
Description of the drawings
Fig. 1 is the flow process of a kind of vehicle-mounted fingerprint identification method fingerprint digital picture pretreatment of the invention and Finger print characteristic abstract
Figure；
Fig. 2 is the situation for being currently needed for the pixel that processes in one embodiment of the invention under different eight neighborhood qualificationss
Schematic diagram；
Fig. 3 is that a kind of vehicle-mounted fingerprint identification method fingerprint digital picture of the present invention puts flow chart in storage；
Fig. 4 is that a kind of vehicle-mounted fingerprint identification method fingerprint digital picture of the present invention matches flow chart.
Fig. 5 is the flow chart for carrying out vehicle-mounted fingerprint recognition using the present invention.
Specific embodiment
The present invention is described in detail for shown each embodiment below in conjunction with the accompanying drawings, but it should explanation, these
Embodiment not limitation of the present invention, those of ordinary skill in the art according to these embodiment institute work energy, method,
Or the equivalent transformation in structure or replacement, belong within protection scope of the present invention.
Shown in ginseng Fig. 1, a kind of vehicle-mounted fingerprint identification method is present embodiments provided, including：Fingerprint collecting, fingerprint digitized map
As pretreatment, Finger print characteristic abstract and fingerprint minutiae matching；Wherein, fingerprint digital picture pretreatment includes：Image sharpening；Gray scale
Conversion；Image binaryzation；Image thinning；Finger print characteristic abstract includes Endpoint ID；Bifurcated point identification；Remove dummy results.
In the present embodiment, fingerprint digital picture pretreatment also includes：Fingerprint image gray scale normalization；Fingerprint image gray scale
Equalization；Fingerprint Image Segmentation；Gabor is filtered.
In the present embodiment, image sharpening includes：Wave filter is generated using fspecial functions, imfilter letters are recycled
It is several that convolutional calculation is carried out to image；Wherein fspecial type function is unsharp types.
In the present embodiment, image thinning includes：Image is changed into into single pixel connected graph by the way of template matching.
In the present embodiment, removing dummy results includes：The point of image border is cut off using the method that fingerprint image cuts；
Characteristic point closer to the distance is removed using distance threshold method.
In the present embodiment, Fingerprint Image Segmentation is using the fingerprint segmentation method based on gradient.
In the present embodiment, fingerprint minutiae matching includes：The intermediate point of fingerprint digital picture to be matched is extracted, in the middle of this
Point is used as matching reference minutiae；Using end points and the characteristic vector template of intersection feature construction warehouse-in fingerprint；By matching reference minutiae
Contrasted with the match point in template base, fingerprint similarity is weighed using Euclidean distance, calculated whether fingerprint matches.
Below the present invention is elaborated.
1st, Pre-processing Algorithm of Fingerprint Recognition.
With the continuous development of Digital Image Processing and hardware technology, the acquisition of fingerprint is by the original side for restraining ink
Formula has turned to the advanced Acquisition Instrument by light electrosynthesis to gather, and is then converted into digital picture, then carries out on computers
Process, so as to obtain the characteristic information of fingerprint.It is special due to restraining firmly uneven, surface skin during fingerprint image acquisition
Property, the impact of a variety of causes such as acquisition condition, imaging sensor, the fingerprint image of collection is one secondary containing various different degrees of noises
The gray level image of interference, streakline there may be fracture, adhesion or the phenomenon such as smudgy.The presence of these noises will have a strong impact on
The accuracy of fingerprint recognition, then this is accomplished by the fingerprint digital picture first to gathering and carries out first step operation diagram picture
Pretreatment.
In the present invention, preprocessing part includes fingerprint image gray scale normalization and equalization, Fingerprint Image Segmentation, filters, refers to
Print image greyscale transformation, fingerprint image sharpening, the refinement of fingerprint image black white binarization, fingerprint image.
1) fingerprint image gray scale normalization and equalization
Fingerprint image gray scale normalization
Normalized purpose is to eliminate during fingerprint collecting due to the noise of sensor itself and because finger pressure
The gray difference that power is different and causes, the contrast and gray scale of fingerprint image is adjusted in a fixed rank, is follow-up
Process and a more unified picture specification is provided.Can be normalized according to the following formula：
Wherein, I (i, j) is the gray value of point (i, j), Mean, VARIt is gray value and the variance of original image, M0、It is the phase
The gray average of prestige and variance.
The image discrete for one, i-th gray level riFrequency of occurrences number niRepresent, the gray-scale pixels correspondence
Probit be：
In formula：N is sum of all pixels, riThe gray value of original image, meets normalizing condition；pr(ri) it is probit.
Fingerprint image grayscale equalization
Equalization is to enter line broadening to the gray level more than number of pixels in image, and the gray level few to number of pixels contracts
Subtract.The function expression equalized by image is:
Wherein, k is gray level, T (ri) be transforming function transformation function, SiFor the pixel value r of original imageiCorresponding value.Accordingly
Contravariant is changed to：
Si=T-1(Si) (4)
2) Fingerprint Image Segmentation
The target of Fingerprint Image Segmentation is exactly the needs according to feature extraction, poor quality in fingerprint image, follow-up
Image-region and the effective coverage for being difficult to recover in process separates, and enables subsequent treatment to concentrate on effective coverage.Can improve special
The degree of accuracy of extraction is levied, the time of fingerprint pretreatment can be greatly reduced.Fingerprint segmentation algorithms most in use have based on gray variance method and
Fingerprint segmentation method based on direction.The present invention is using the fingerprint segmentation method based on gradient.
Method based on gradient typically with the concordance of gradient as feature come Segmentation Algorithm of Fingerprint image, because of fingerprint effective coverage
It is parallel linear structure, so the concordance of foreground area is typically all higher than background area.The concordance put at (i, j) can be with
Definition is as shown in formula (5).
In formula, Vx, VyRepresented in G (i, j) point x, the gradient in y directions respectively.
The directional information of fingerprint image is taken full advantage of based on the dividing method of gradient, but it is again unlike dividing based on direction
The result that algorithm utilization orientation direct like that is calculated is cut, but make use of the concordance of gradient, solve segmentation other side well
To heavy dependence.
3) Gabor filtering
Gabor filtering using nonlinear smoothing device, have excellent nonlinear smoothing device performance and with biological vision system
Close the characteristics of.
Gabor nonlinear smoothing device general types are：
In formula,It is the orientation of Gabor nonlinear smoothing devices, f is the frequency of nonlinear smoothing device, δxAnd δyIt is along X respectively
Axle and Y-axis Gaussian envelope constant.The texture knot of the crestal line and valley line in specified orientation can be strengthened using suitable parameter
The information such as structure, the mixed and disorderly digital noise of suppression.The feature representation obtained after Gabor nonlinear smoothings effectively can be represented
Fingerprint digital picture, so the selection of parameter just becomes the committed step of construction Gabor nonlinear smoothing devices.On the digital image
The frequency of nonlinear smoothing device, orientation, three ginsengs of Gaussian envelopes constant be must determine using Gabor nonlinear smoothing devices
Number.Frequency characteristic f of nonlinear smoothing device is determined that orientation is also by the orientation of certain block scope completely by the frequency of certain block scope
Determine.δxAnd δyPeek value needs depend on the circumstances.Peek value is bigger, stronger to digital noise adaptability, it is likely that producing
Raw pseudo- streakline；Numerical value is less, and the probability for producing pseudo- details is less, but the ability of removal digital noise will weaken.
4) fingerprint image greyscale transformation (image enhaucament)
What the present embodiment was adopted is exactly to use histogram equalization to strengthen gray level image contrast.Histogram modification is referred to
Strengthen image pixel value histogram distribution to carry out enhancement process to image, after histogram modification, image pixel value is each
All it is distributed in individual rank, the margi n pulls big between image pixel value, it is easier to show image detail.
In digital picture, gray level is rkPixel occur frequency (probability), in formula (9) N for piece image total pixel
Number, nkIt is the pixel count of kth level gray scale, rkRepresent k-th gray level, P (rK) represent the relative frequency that the gray level occurs.
When carrying out histogrammic equalization to the digital picture, corresponding discrete form is：
In formula, S is uniformly distributed into S for r distribution transformations.
Greyscale transformation is done using formula (10), you can obtain the image after histogram equalization.Image histogram typically can be from
Piece image is described on the whole, such as original image is under-exposed or over-exposed, it is also possible to find out dynamic range select it is suitable with
It is no.The rectangular histogram of image can be modified, to reach the purpose for improving image.
1 histeq functions of table
Histogram modification generally has two big class of histogram equalization and histogram specification.What is adopted in the present embodiment is exactly
Histogram equalization.The histogram modification method commonly used during histogram equalization, can generally obtain satisfied image
Effect, as shown in table 1.
5) fingerprint image sharpens (image enhaucament)
For strengthening the boundary between fingerprint ridge, projecting edge information is beneficial to binaryzation, and Edge contrast is necessary.It is sharp
Change and process for enhanced contrast and detect that edge is very useful.The effect of image sharpening is exactly the profile for compensating image, makes figure
As more clear.The present embodiment generates wave filter using fspecial functions, recycles imfilter function pair images to be rolled up
Product calculates the effect so as to reach image sharpening.Wherein fspecial type function is unsharp types, is strengthened for contrast
Wave filter, parameter alpha are used for controlling the shape of wave filter, and scope is [0,1], and default value is 0.2.
6) fingerprint image black white binarization
A part of the fingerprint digital picture binaryzation as fingerprint digital picture preprocessing process, is to carry out fingerprint digitized map
As the basis of micronization processes.Bianry image refers to only black (gray value is 0) white (gray value is 1) two-value in entire image picture
Image, does not present the change of gray scale on them.In Digital Image Processing, binary image occupies very importantly
Position.This is because in practical image processing system, it is desirable to which the speed of process is high, low cost, the shading image for containing much information
Process cost too big, be not very wise move.And the image after binaryzation can be analyzed with the concept in geometry and be retouched with feature
State, it is convenient many compared with for gray level image.Thus binary image processes the independence for having become in image procossing at present
, important branch and be widely applied.For the identification of fingerprint, some information must include the two of crestal line and valley line
In value description.Thus must determine according to original gray level image that the every bit on image should belong to object region, so as to produce
The corresponding bianry image of life.It not only can greatly reduce amount of storage, and to causing differentiation process to be below disturbed less, greatly
The big processing method for simplifying thereafter.Image after binaryzation is the basis with post processing, and its algorithm has directly to process below
The impact for connecing.
7) fingerprint image refinement
Process of the shape information of object to fingerprint digital picture is particularly significant.Refinement is before the characteristic point that takes the fingerprint
Last one procedure.Briefly, refinement is, under the topological connection relation for not affecting artwork, the width of ridge to be reduced to single
The processing procedure of pixel wide.One good thinning method is to maintain the seriality of original ridge, reduces as anthropic factor is made
Into impact.Situations such as anthropic factor main jagged and short crestal line, these have in all causing the feature for extracting a lot
Pseudo-characteristic.The advantage of refinement is to reduce memory headroom, and it only needs to required structural information in storage image.So, right
Data structure can be simplified in the process of image.
Refinement structure is different and variant according to refinement mode.In the same of the Topology connection for not changing original image
When, the structure of refinement should be strict eight neighborhood digital picture skeleton；Remove in streakline beyond characteristic point, each pixel only with
Two neighboring pixel is eight neighborhood, erases any pixel all by the connectivity of destruction streakline.Streakline micronization processes process will meet
Convergence, topological, connectivity, refinement property, axis, retentivity, rapidity etc. are required.
Mostly the comparison of refinement mode is the mode of template matching.This mode is the local neighborhood according to certain pixel
The digital picture feature of (such as 3 × 3,5 × 5 etc.) is processed to which, and such as iterative method, OPTA simply connected methods etc. are all to adopt
The method of template matching.The method of refinement is also using refinement sides such as Edge Search coding, outline calculating and neutral nets
Formula.
For the scope of arbitrary shape, in thinning process, this to be judged according to the situation of the eight of each pixel consecutive points
Whether point can reject or retain, and which is substantially similar to etching operation, for example, give the pixel for being currently needed for processing in Fig. 2
Situation under different eight neighborhood qualificationss.
As can be seen from Figure 2：A () figure can not be removed, because it is an internal point, we are required of skeleton, if even
Internal point is also removed, and skeleton can also be emptied；It is identical with (a), can not (b) remove；C () can remove, such point is not
It is skeleton；D () can not remove, because after removing, originally connected part is disconnected；E () can remove, such point is not
Skeleton；F () can not remove, because it is the end points of straight line, if such point is removed, then last whole straight line is also moved
Except what does not remain；G () can not remove, because the skeleton of isolated point is exactly its own.According to Fig. 2, we can summarize
Such as draw a conclusion：Internal point, isolated point and straight line end points can not all be removed；When P is boundary point, if not increasing company after removing P
Reduction of fractions to a common denominator amount can then remove P points.
The refinement purpose of fingerprint image is that image is changed into single pixel connected graph, and can the quality of thinning effect is directly affected
Minutia is extracted accurately.
2nd, Finger print characteristic abstract
After the binaryzation digital picture through refining is obtained, then seek to carry out key to the digital picture after refinement
Feature extraction, so that reach the purpose of the different fingerprint digital picture of identification.Feature extraction is referred to the stricture of vagina of fingerprint digital picture
The form of the feature numerical value such as line trend, end points, cross point shows such that it is able to sufficiently represent the unique of fingerprint image
Property, i.e., the process of global characteristics is extracted from pretreated fingerprint image.
The method of fingerprint digital picture feature extraction mainly has two kinds, and one is recognized carefully from the former gray level image of fingerprint
Section feature, one is from the refined image of fingerprint to recognize minutia.The feature for taking the fingerprint mainly has two classes, one of them
It is the category feature of fingerprint, i.e. global characteristics, another is the local feature of fingerprint, and it includes end points, bifurcation, bifurcation point, orphan
The information such as vertical point.
What the present embodiment was to be extracted is end points and bifurcation information.Lower mask body introduce the extraction of end points and bifurcation and
How dummy results are removed.
1) extraction of end points and bifurcation
End points：One fingerprint circuit terminates in the point.
Bifurcation：One fingerprint circuit here is separated into the fingerprint circuit of two or more.
Feature on fingerprint except with classification characteristic in addition to, also with three below different qualities：
(1) direction character point can be towards certain direction；
(2) speed that the direction of curvature representation lines changes；
(3) position of position feature point can with coordinate (x, y) representing, it can be it is absolute, or
Relative triangulation point or characteristic point.
2) dummy results are removed
The characteristic point of many falsenesses occurs in fingerprint feature point extraction process, it is therefore desirable to removed, to ensure to refer to
The effectiveness of stricture of vagina feature.Fingerprint characteristic goes pseudo-operation mainly to filter out the characteristic point for not meeting fingerprint characteristic.Pseudo-characteristic has
Following characteristics：It is mostly in image border；Pseudo-random numbers generation inside image is closer to the distance, and two or more pseudo-characteristics are simultaneously
It is present in the region of very little.The present embodiment proposes two kinds according to these features and removes fake method：Firstly for image border
Point, the method cut using fingerprint image, i.e., directly cut away to the point at edge；Then using distance threshold method remove distance compared with
Near characteristic point.
3rd, fingerprint image matching
Minutia of the characteristic matching of fingerprint digital picture mainly to being extracted is matched, and what is will compared is new
The minutia numerical value of the image in characteristics of image numerical value and the fingerprint base set up is compared, and by image the most similar
Export as a result, this is the checking identification process of fingerprint digital picture, and the final purpose of fingerprint recognition system.
1) arrange threshold value to be matched
Before fingerprint digital picture is matched, the intermediate point of fingerprint digital picture to be first extracted, feature templates are resettled,
These are all to carry out matching the preparation done for fingerprint digital picture.
(1) point location in the middle of
Intermediate point refers to the maximum point of the ridge curvature of curve of fingerprint digital picture, in the algorithm of the present embodiment, by centre
Point is used as matching reference minutiae.Here the centre chosen is a small range in the middle of fingerprint digital picture, first obtains fingerprint numeral
The point orientation of image, the mean values of adjacent 8 gray values sums, then this 8 gray values are sought with the difference of mean values
With the orientation at minimum place is the orientation of dictionary place fingerprint digital picture crestal line, so as to obtain an orientation diagram.Point orientation
Figure is divided into the fritter of 16 × 16 sizes, and to calculating rectangular histogram per block, its peak number value orientation is block orientation, i.e., the master at every piece midpoint
Lead orientation.Then on this thick block orientation diagram according to following principle removal search intermediate range, reviewing party's block number group line by line.So
Afterwards further according to the angle for obtaining each orientation and the mean values of adjacent 8 gray values sums, then seek this 8 gray values
It is difference with mean values and, the orientation at minimum place is the orientation of this point place fingerprint digital picture crestal line, thus
Point orientation diagram is arrived.
(2) feature templates are set up
End points and cross point are the principal characters of fingerprint refined image, and we can be constructed using both principal characters
The characteristic vector template of fingerprint.Feature end points is categorized as 1, and feature bifurcation is categorized as 2；Feature end points is set up with respect to intermediate point
Distance vector, and feature bifurcation is with respect to the distance vector of intermediate point；Position vector of the feature end points with respect to intermediate point is set up, and
Position vector of the feature bifurcation with respect to intermediate point.Feature extracting method is as follows：
If Cn(P) it is crossing number, is Sn(P) pixel, 8- neighborhoods
For a width thoroughly fingerprint image of refinement, only three kinds streaklines：(1)Cn(P)=1, Sn(P)=1 referred to as holds
Point；(2)Cn(P)=2, Sn(P)=2,3,4 are referred to as continuity point；(3)Cn(P)=3, Sn(P)=3 are referred to as crunode.If the spy for extracting
Levy point set P to represent, the number of characteristic points of the wherein n by extracting, Pi=(Xi,Yi,Ti,Ai) represent characteristic point coordinate；Ti
The type of characteristic point is represented, the T when characteristic point is end pointsi=2；Ai represents the angle of characteristic point, and the angle of end points is taken from end points
For the angle of the end line of starting point.The angle of end line and branch asks the method to be：Start search continuity point from characteristic point another until searching out
Individual characteristic point or step-length reach 7, if the last point for searching is (X, Y), have：
(3) define match point
Defining match point PointOfModel (characteristic point in point set P) is extracted from the fingerprint digital picture of input
Out, another match point PointOfMatch (characteristic point in point set Q) is carried from fingerprint digital picture storehouse
Take out and be stored in template base, two groups of point sets are compared.
2nd, the matching algorithm of fingerprint digital picture
Fingerprint digital picture matching is always a difficult problem in fingerprint digital picture Model Identification.It is contained for two
There are the point set P { p of varying number1,p2,...pmAnd Q { q1,q2...qnHow to find out between them matching association.Accordingly, it is capable to
The fingerprint digital picture matching algorithm that the algorithm of the geometry variable problem between two point sets is only is efficiently solved enough.Fingerprint number
Word images match presently, there are many algorithms, and such as relaxed algorithm, the document wherein having only have been processed between point model because of smooth belt
The deviation come；Also the streakline having has been processed because smoothing and turning the error brought.
Two point sets P and Q in fingerprint digital picture matching, wherein P are extracted from the first width digital picture, have m feature
Point is constituted, and Q is extracted from the second width digital picture, has n characteristic point to constitute, i.e. P { p1,p2,...pmAnd Q { q1,q2...qn}.Cause
For in actual applications, the relative position of point has error, so the matching between them is exactly that each characteristic point is distinguished
Extract, then the process that the vector obtained relative to intermediate point is compared, two point sets is had between maximum quantity point pair and exist
Stable one-to-one corresponding association.
For a characteristic point concentrated, the present embodiment are described with the coordinate in x orientation and y orientation, fingerprint digital picture is every
One characteristic point is five dimensional vector (x, y, β, t, c), and wherein x, y are the coordinate position of the point respectively, β be characteristic point relative to
The position vector of intermediate point, t are characterized type (being bifurcation or end points) a little, and c is characterized a little relative to intermediate point
Distance vector.
Therefore some suitable algorithms, by searching in a certain parameter space, how many fingerprint digital picture can be used
To matching, i.e. Matching supporting number.When the Matching supporting number of gained is maximum, the result required for also just having obtained.The present embodiment
Fingerprint similarity is weighed using Euclidean distance, so as to calculate whether fingerprint matches.
Euclidean distance (Euclidean distance) is also referred to as Euclidean distance, and it is that a distance for generally adopting is fixed
Justice, it is the actual distance in n-dimensional space between two points.
What is referred in the Euclidean distance in two and three dimensions space is exactly distance between 2 points.
Two dimension formula be：D=sqrt ((x1-x2)2+(y1-y2)2) (14)
Three-dimensional formula is：D=sqrt ((x1-x2)2+(y1-y2)2+(z1-z2)2) (15)
Certainly Euclidean distance can also be generalized to n-dimensional space, and the formula of the Euclidean distance of n-dimensional space is：
Here i=1,2 n.Xi1 represents the i-th dimension coordinate of first point.N dimension theorem in Euclid space is a point set, it
Each point can represent (x (1), x (2), x (n)), and wherein x (i) (i=1,2 n) is real number, referred to as the i-th of x
The distance between individual coordinate, two point x and y=(y (1), y (2) y (n)) d (x, y) is defined as above formula.
Briefly, two stack features vectors are exactly subtracted each other by Euclidean distance, then obtain putting down for the difference between their correspondences
Fang He, then opens radical sign.For example：Then the distance between they are exactly d=sqrt ((1-4) to A=(1,2,3) B=(4,5,6)2+
(2-5)2+(3-6)2), then system just searches for database file one by one, seeks the minima of their distances.If between them
Distance is 0, then system thinks that this two digital pictures is the same, comes from same finger in other words.Euclidean distance regards letter as
Number similarity degree, distance it is more near more similar, more easily interfere, the bit error rate is higher.
Below by instantiation, the invention will be further described：
Shown in ginseng Fig. 3 and Fig. 4.
(1) select fingerprint digital picture
It is required that selected fingerprint image must be black and white, and must be 256 × 256 sizes.Meeting these will
Following operation can just be carried out after asking, point out always to select black white image if the requirement is unsatisfactory for.
(2) fingerprint image preprocessing
Preprocessing part includes：The determination of Core Point in Fingerprint, fingerprint image gray scale normalization and equalization, fingerprint image
Picture segmentation, Gabor filtering, fingerprint image sharpening, fingerprint image greyscale transformation, fingerprint image black white binarization, fingerprint image are thin
Change.Causing through a series of preprocessing process originally may fingerprint digital picture change score that is unintelligible or having other defect
It is bright, it is that feature extraction below is had laid a good foundation.
(3) fingerprint image characteristics are extracted and are put in storage
After filtering through Gabor, in an orientation nonlinear smoothing, corresponding those partial fingerprints in ridge orientation
The discrimination of the ridge of digital picture and paddy is just strengthened, and can thus reduce the error of feature extraction, improves system fingerprint
The degree of accuracy of Image-matching.
The present embodiment employs 64 secter pats and processes one by one, and Program extraction is each fingerprint digital picture gray values
Absolute index it is poor, feature extraction algorithm step is as follows in this system：First have to determine the intermediate point of fingerprint digital picture；
Secondly according to intermediate point, fingerprint digital picture is cut, system is divided into some fritters it, in this system, divide into 64
Fritter；Finally with eight orientation nonlinear smoothing devices, make the orientation of ridge just be strengthened, an orientation non-thread is obtained through processing
The mild-natured characteristic vector for sliding onto, system are stored in this characteristic vector in data base the foundation as matching.
(4) fingerprint is selected with fingerprint matching in storehouse
The searching algorithm that system is used is simple sequential search mode, is exactly the search one by one to database file, is
When system reads in a fingerprint digital picture picture, just the mode according to more than, carries out feature to this width fingerprint digital picture and carries
Take, encode, be saved in inside temporary variable and go.One group fingerprint is also preserved in the fingerprint digital picture database file of system equally
The feature of digital image gray level deviation.Detailed process is as follows：
Apply for memory headroom first, preserve the temporary variable of matching result, and preserve the fingerprint to be matched to system input
The coding of digital picture；Then the database file of search system, if database file is sky, terminates to search；Finally, if being
The database file of system is not space-time, due to when fingerprint digital picture is put in storage, to fingerprint numeral inside database file
The certain order of picture coding, equally, system is also to the order that fingerprint digital picture encoding setting to be matched is the same.So it is
System can be matched successively, so can judge the distance between two fingerprint digital pictures according to Euclidean distance matching algorithm.
It is previously noted system to encode a fingerprint digital picture preservation twice, this system is exactly to compare corresponding feature, is taken
Wherein less one result compared as system.
Herein using by comparing image to be tested with the Euclidean distance of template image correlated characteristic point in data base come real
Existing, a fingerprint image, if match index is more than definite value T, then it is assumed that the match is successful, otherwise fails.
(5) output result
Matching result is shown with message box, the Euclidean distance numbered including fingerprint in matching library and calculate.
Resume module process main in the present embodiment includes：
(1) input picture
Use in the present embodiment from drop-down menu select image mode, GUI design used in be
popupmenu.Call back function arranges 6 patterns, reads in corresponding picture when selection pattern is a few behind.
It should be noted that the display problem after reading.As the present embodiment has the process of image comparison, so needing
Two viewing areas are wanted, this is accomplished by the two is respectively displayed on different regions when picture is read in.Even if using sentence axes
(handles.axes1) it is displayed in axes1.Merely the sentence can not be write on before imshow in embodiment, but
Before writing on selection pattern, this is only real realization and two width fingerprint images is respectively displayed in two different regions.
(2) image enhancement processing
In order to ensure the working effect of automatic system of fingerprint recognition, need to carry out image enhancement processing to fingerprint image.Increase
It is strong mainly to include two parts：Sharpen and greyscale transformation.Sharpening seeks to first generate wave filter fspecial, recycles imfilter
Carry out image convolution calculating；What the numerical value of gray level image was represented is the brightness of gray level image, and the luminance contrast of gray level image can be with
Represented using contrast, contrast can be expressed as：Contrast=brightness maxima/minima.
If contrast is less, it is meant that the maximum of brightness and minima are more or less the same, the contrast of image is not strong, figure
As content difference less, the content of image is not seen；If contrast is larger, it is meant that the maximum of brightness and minima difference
Than larger, the contrast of image is stronger, and the content difference of image can see the content of image than larger.
Strengthening picture contrast really strengthens each several part contrast of original image, that is to say, that interested in enhancing image
Gray areas, it is relative to suppress those uninterested gray areas.Used in the present embodiment increased using histogram equalization
Strong lime degree picture contrast.
(3) image black white binarization
Black white binarization process is carried out to image.After treatment, the carina in fingerprint becomes apparent black line, plough
Ditch becomes white.
(4) image thinning
The carina of fingerprint is refined.Carina refinement eliminates the unnecessary pixel in carina both sides, and the width of last carina becomes
One pixel width.
(5) Endpoint ID
A.Minutie is filtered
Process is filtered to carina using wave filter minutie.Minutie wave filter will be using 3 × 3 windows at
Reason image pixel.If central pixel point be 1 and only one of which value be 1 neighborhood territory pixel point, central pixel point is carina end
Point.If center pixel be 1 and surrounding have 3 values for 1 neighborhood territory pixel point, central pixel point is a bifurcation.If
Central pixel point be 1 and surrounding have 2 values for 1 neighborhood territory pixel point, then central pixel point is generic pixel point.
B. Endpoint ID is processed
The end points of carina is identified and mark point is drawn.Step is divided into the following steps：
Step 1：Function handle is obtained, minutie functions are used herein；
Step 2：It is filtered process；
Step 3：Select end points and opposite end point is identified；
Step 4：End region is analyzed, end region center is found；
Step 5：End points rounding is integer；
Step 6：Carina end points is identified with red circle.
C. bifurcated point identification
The bifurcation of carina is identified
D. dummy results are removed
The falseness mark to producing in carina end points and bifurcation identification procedure is removed, in order to remove dummy results, is processed
Process is as follows：
Process 1：If the distance between an end points and a bifurcation are less than D, remove this result.
Process 2：If the distance between two bifurcations are less than D, remove this result.
Process 3：If the distance between two end points are less than D, remove this result.
(6) set image processing region and show details
Need to determine an image processing region ROI.In order to realize the determination of ROI, it is considered to use black white image, while to figure
As using a closed operation and an erosion operation.Transparency is set and Endpoint ID and bifurcated point identification is drawn, to mark
Figure and ROI figures are overlapped display.Can also show in artwork, the present embodiment selects mark to show in artwork.
(7) characteristic details are preserved
Characteristic details include carina orientation, carina end points orientation and carina bifurcated orientation, and its median carina orientation is to determine
After different details, the orientative feature of each details can be found；Carina end points orientation is to find end points orientation, in order to realize
End points orientation is searched, and using the borderline shop of 5 × 5 end point analysis window analysis, the position for obtaining is carried out process of tabling look-up.
The purpose of the system is that a kind of fusion fingerprint digital picture feature of offer and minutiae feature are matched, so as to
Quickly realize that fingerprint digital picture is known otherwise, step is as follows：
The first step, reference point detection and its orientation determination, i.e., carry out digital picture distortion correction, number to input digital image
After word image segmentation, orientation field is smoothed according to the concordance in different size scope inside gradient orientation, and utilizes multiresolution model
The detection of attribute for enclosing Orientation differences goes out the reference point of fingerprint digital picture, and then with reference point as centre, observes different size
In fan-shaped range, the orientation of fingerprint digital picture and the difference of radial orientation are so that it is determined that go out the reference azimuth of reference point.
Second step, digital picture feature extraction is centre to fingerprint digital picture sectional drawing, sectional drawing that is, with the reference point for detecting
Size 128 × 128, is the digital picture feature for obtaining fingerprint digital picture, carries out two grades of wavelet transformations first to digital picture,
For low frequency part carries out Fourier transformation again, i.e., wavelet transformation is first carried out to low frequency part, it is extremely right after highpass non-linear is smooth
Number coordinate transform, carries out wavelet transformation afterwards again, obtains 64 × 64 coefficient matrixes only, carry out feature choosing to which after standardization
Select, finally give the characteristic vector of fingerprint digital picture；
3rd step, details point template are set up, i.e., the sectional drawing for obtaining to second step carries out digital image enhancement, binaryzation, thin
After the operation such as change, the refinement digital picture of fingerprint digital picture sectional drawing is obtained, the minutiae point of fingerprint digital picture is then extracted,
Reference point with detection is transformed into the minutiae point for obtaining under polar coordinate system as limit, thin so as to set up fingerprint digital picture
Node template；
4th step, characteristic matching, i.e., to the fingerprint digital picture to be compared being input into, the finger obtained using above three step
Stricture of vagina digital picture feature and minutiae template judge whether two pieces of fingerprint digital pictures match.
5th step, the match is successful.
Shown in ginseng Fig. 5, Fig. 5 is the flow chart for carrying out vehicle-mounted fingerprint recognition using the present invention.
Key is firstly inserted into, activate switch is beaten to ON shelves, fingerprint recognition system is powered, car owner's input fingerprint, Jing data
Storehouse returns matching result.If fingerprint matching success, fingerprint recognition system control starts the control circuit closure of motor and oil pump
So that circuit turn-on, then beats activate switch to START shelves, automobile normally starts；If fingerprint matching is unsuccessful, start motor
(simulation of Single-chip Controlling relay) is disconnected still with the control circuit of oil pump, car owner re-enters fingerprint, if three inputs refer to
Stricture of vagina it fails to match then start report to the police.
The a series of detailed description in detail of those listed above is only for feasibility embodiment of the invention specifically
Bright, they simultaneously are not used to limit the scope of the invention, all equivalent implementations made without departing from skill spirit of the present invention
Or change should be included within the scope of the present invention.
It is obvious to a person skilled in the art that the invention is not restricted to the details of above-mentioned one exemplary embodiment, Er Qie
In the case of spirit or essential attributes without departing substantially from the present invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, embodiment all should be regarded as exemplary, and be nonrestrictive, the scope of the present invention is by appended power
Profit is required rather than described above is limited, it is intended that all in the implication and scope of the equivalency of claim by falling
Change is included in the present invention.
Claims (7)
1. a kind of vehicle-mounted fingerprint identification method, including：Fingerprint collecting, fingerprint digital picture pretreatment, Finger print characteristic abstract and refer to
Stricture of vagina characteristic matching；Wherein, the fingerprint digital picture pretreatment includes：Image sharpening；Greyscale transformation；Image binaryzation；Image
Refinement；The Finger print characteristic abstract includes Endpoint ID；Bifurcated point identification；Remove dummy results.
2. a kind of vehicle-mounted fingerprint identification method according to claim 1, it is characterised in that the fingerprint digital picture is located in advance
Reason also includes：Fingerprint image gray scale normalization；Fingerprint image grayscale equalization；Fingerprint Image Segmentation；Gabor is filtered.
3. a kind of vehicle-mounted fingerprint identification method according to claim 1, it is characterised in that described image is sharpened to be included：Profit
Wave filter is generated with fspecial functions, recycles imfilter function pairs image to carry out convolutional calculation；It is wherein described
Fspecial type function is unsharp types.
4. a kind of vehicle-mounted fingerprint identification method according to claim 1, it is characterised in that described image refinement includes：Adopt
Image is changed into into single pixel connected graph with the mode of template matching.
5. a kind of vehicle-mounted fingerprint identification method according to claim 1, it is characterised in that described to remove dummy results bag
Include：The point of image border is cut off using the method that fingerprint image cuts；Characteristic point closer to the distance is removed using distance threshold method.
6. a kind of vehicle-mounted fingerprint identification method according to claim 2, it is characterised in that the Fingerprint Image Segmentation is adopted
Fingerprint segmentation method based on gradient.
7. a kind of vehicle-mounted fingerprint identification method according to claim 1, it is characterised in that the fingerprint minutiae matching bag
Include：The intermediate point of fingerprint digital picture to be matched is extracted, using the intermediate point as matching reference minutiae；Using end points and cross point
Latent structure puts the characteristic vector template of fingerprint in storage；The matching reference minutiae is contrasted with the match point in template base, profit
Fingerprint similarity is weighed with Euclidean distance, calculates whether fingerprint matches.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610880701.8A CN106529407A (en) | 2016-10-09 | 2016-10-09 | Vehicle-mounted fingerprint recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610880701.8A CN106529407A (en) | 2016-10-09 | 2016-10-09 | Vehicle-mounted fingerprint recognition method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106529407A true CN106529407A (en) | 2017-03-22 |
Family
ID=58331728
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610880701.8A Pending CN106529407A (en) | 2016-10-09 | 2016-10-09 | Vehicle-mounted fingerprint recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106529407A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109540918A (en) * | 2018-11-28 | 2019-03-29 | 鞍钢集团自动化有限公司 | A kind of hot rolled coil edge fault detection device and method |
| CN109840487A (en) * | 2019-01-24 | 2019-06-04 | 易保互联医疗信息科技（北京）有限公司 | The private key generation method and system of block chain stored value card based on finger print information |
| WO2019237520A1 (en) * | 2018-06-11 | 2019-12-19 | 平安科技（深圳）有限公司 | Image matching method and apparatus, computer device, and storage medium |
| CN110920507A (en) * | 2019-11-07 | 2020-03-27 | 张彩萍 | Car as a house |
| CN111553233A (en) * | 2020-04-22 | 2020-08-18 | 广东电网有限责任公司 | Intelligent vehicle management method |
| CN113343963A (en) * | 2021-08-09 | 2021-09-03 | 北京惠朗时代科技有限公司 | Intelligent trunk fingerprint identification method and system based on high-quality template |
| WO2023151337A1 (en) * | 2022-02-09 | 2023-08-17 | 腾讯科技（深圳）有限公司 | Identity recognition method and apparatus, and feature extraction method and apparatus for biological texture information |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101609499A (en) * | 2008-06-20 | 2009-12-23 | 南京理工大学 | Quick fingerprint identification method |
| CN105320946A (en) * | 2015-11-03 | 2016-02-10 | 盐城工学院 | MATLAB based fingerprint identification method |
-
2016
- 2016-10-09 CN CN201610880701.8A patent/CN106529407A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101609499A (en) * | 2008-06-20 | 2009-12-23 | 南京理工大学 | Quick fingerprint identification method |
| CN105320946A (en) * | 2015-11-03 | 2016-02-10 | 盐城工学院 | MATLAB based fingerprint identification method |
Non-Patent Citations (7)
| Title |
|---|
| 冯伟兴 等: "《Visual C++数字图像模式识别典型案例详解》", 31 July 2012 * |
| 卢文祥 等: "《机械工程测试.信息.信号分析》", 31 May 2014 * |
| 李华 等: "《计算机辅助设计与图形学新进展》", 31 October 2006 * |
| 王可洲 等: "《汽车电气设备构造与维修》", 31 July 2014 * |
| 王威 等: "《指纹信息工作教程》", 31 May 2012 * |
| 耿德英: "指纹识别中图像处理方法的设计与实现", 《万方学位论文数据库》 * |
| 郑继刚 等: "《基于MATLAB的数字图像处理研究》", 31 December 2010 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019237520A1 (en) * | 2018-06-11 | 2019-12-19 | 平安科技（深圳）有限公司 | Image matching method and apparatus, computer device, and storage medium |
| CN109540918A (en) * | 2018-11-28 | 2019-03-29 | 鞍钢集团自动化有限公司 | A kind of hot rolled coil edge fault detection device and method |
| CN109540918B (en) * | 2018-11-28 | 2021-04-16 | 鞍钢集团自动化有限公司 | Hot-rolled coil edge defect detection device and method |
| CN109840487A (en) * | 2019-01-24 | 2019-06-04 | 易保互联医疗信息科技（北京）有限公司 | The private key generation method and system of block chain stored value card based on finger print information |
| CN109840487B (en) * | 2019-01-24 | 2021-03-23 | 易保互联医疗信息科技（北京）有限公司 | Private key generation method and system of block chain electronic wallet based on fingerprint information |
| CN110920507A (en) * | 2019-11-07 | 2020-03-27 | 张彩萍 | Car as a house |
| CN111553233A (en) * | 2020-04-22 | 2020-08-18 | 广东电网有限责任公司 | Intelligent vehicle management method |
| CN113343963A (en) * | 2021-08-09 | 2021-09-03 | 北京惠朗时代科技有限公司 | Intelligent trunk fingerprint identification method and system based on high-quality template |
| WO2023151337A1 (en) * | 2022-02-09 | 2023-08-17 | 腾讯科技（深圳）有限公司 | Identity recognition method and apparatus, and feature extraction method and apparatus for biological texture information |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106529407A (en) | Vehicle-mounted fingerprint recognition method | |
| Kuruvilla et al. | A review on image processing and image segmentation | |
| CN100412883C (en) | Fingerprint identifying method and system | |
| CN103886589B (en) | Object-oriented automated high-precision edge extracting method | |
| CN101142584B (en) | Method for facial features detection | |
| CN107122777A (en) | A kind of vehicle analysis system and analysis method based on video file | |
| CN108197644A (en) | A kind of image-recognizing method and device | |
| CN101859382A (en) | License plate detection and identification method based on maximum stable extremal region | |
| CN104751187A (en) | Automatic meter-reading image recognition method | |
| Chatbri et al. | Using scale space filtering to make thinning algorithms robust against noise in sketch images | |
| CN108537143B (en) | A kind of face identification method and system based on key area aspect ratio pair | |
| CN113392856B (en) | Image forgery detection device and method | |
| CN111027544B (en) | MSER license plate positioning method and system based on visual saliency detection | |
| CN111914755A (en) | Eight-direction gradient-solving fingerprint identification model | |
| Mello et al. | Automatic image segmentation of old topographic maps and floor plans | |
| CN111079626B (en) | Living body fingerprint identification method, electronic equipment and computer readable storage medium | |
| CN112686265A (en) | Hierarchic contour extraction-based pictograph segmentation method | |
| Dhar et al. | Bangladeshi license plate recognition using adaboost classifier | |
| CN112200789A (en) | Image identification method and device, electronic equipment and storage medium | |
| CN108921006A (en) | The handwritten signature image true and false identifies method for establishing model and distinguishing method between true and false | |
| CN111612083A (en) | Finger vein identification method, device and equipment | |
| Surmacz et al. | An improved algorithm for feature extraction from a fingerprint fuzzy image | |
| CN113516096B (en) | Finger vein ROI (region of interest) region extraction method and device | |
| Rajithkumar et al. | Template matching method for recognition of stone inscripted Kannada characters of different time frames based on correlation analysis | |
| Dyre et al. | A survey on various approaches to fingerprint matching for personal verification and identification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for 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: 20170322 |