CN107145829A - A kind of vena metacarpea recognition methods for merging textural characteristics and scale invariant feature - Google Patents
A kind of vena metacarpea recognition methods for merging textural characteristics and scale invariant feature Download PDFInfo
- Publication number
- CN107145829A CN107145829A CN201710222874.5A CN201710222874A CN107145829A CN 107145829 A CN107145829 A CN 107145829A CN 201710222874 A CN201710222874 A CN 201710222874A CN 107145829 A CN107145829 A CN 107145829A
- Authority
- CN
- China
- Prior art keywords
- vena metacarpea
- sift
- sift feature
- nbp
- feature point
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 64
- 239000000284 extract Substances 0.000 claims description 6
- 230000027455 binding Effects 0.000 claims description 3
- 238000009739 binding Methods 0.000 claims description 3
- 230000000052 comparative effect Effects 0.000 claims description 2
- 238000004422 calculation algorithm Methods 0.000 abstract description 38
- 238000012216 screening Methods 0.000 abstract description 8
- 239000011159 matrix material Substances 0.000 description 17
- 230000008901 benefit Effects 0.000 description 8
- 210000003462 vein Anatomy 0.000 description 8
- 230000004927 fusion Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000006073 displacement reaction Methods 0.000 description 6
- 230000002708 enhancing effect Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000008030 elimination Effects 0.000 description 4
- 238000003379 elimination reaction Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 239000000047 product Substances 0.000 description 4
- 230000015572 biosynthetic process Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 210000001367 artery Anatomy 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000008570 general process Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 238000012360 testing method 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- 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
- G06V40/14—Vascular patterns
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention discloses a kind of vena metacarpea recognition methods for merging textural characteristics and scale invariant feature, primary dcreening operation is carried out by NBP algorithms, if primary dcreening operation fruiting area is clearly aobvious to have the threshold value t that smallest hamming distance is less than setting1, then corresponding vena metacarpea ROI image is chosen as the recognition result of vena metacarpea ROI image to be identified；Otherwise postsearch screening is carried out using SIFT feature, if aobvious i.e. SIFT feature Point matching logarithm amount is more than or equal to threshold value t clearly for postsearch screening fruiting area2, then it is identified result (identical if there is quantity, what Hamming distance was minimum in them are chosen is used as matching result)；When screen twice substantially can not distinguish result when, then it is assumed that without identification object.The present invention can both lift exclusive use NBP algorithms and the accuracy rate of SIFT algorithms progress vena metacarpea identification reduces misclassification rate, the requirement that collection identification in real time can be met in efficiency again is to have taken into account recognition speed, so overcomes the shortcoming that NBP algorithm misclassification rates are higher and SIFT algorithms are time-consuming.
Description
Technical field
The invention belongs to mode identification technology, more specifically, it is related to a kind of fusion textural characteristics and yardstick not
Become the vena metacarpea recognition methods of feature.
Background technology
Palm vein recognition technical is a kind of biological identification technology emerging in recent years, and its general process includes vena metacarpea
Collection, the pretreatment of image, vein pattern are extracted and characteristic matching identification.According to the difference of feature extracting method, the existing palm
Vein identification method can be attributed to following four classes：It is method based on architectural feature, the method based on subspace, special based on texture
The method of (NBP features) and the method based on scale invariant feature (SIFT) are levied, wherein, (1), the method based on architectural feature
The architectural features such as the line feature or point feature of vena metacarpea are extracted to represent vena metacarpea.Such Method And Principle is simple, relatively more straight
See, but due to easily losing branch, line feature because of vena metacarpea information fuzzy, so that identification capability is limited；(2), based on sub empty
Between method regard vein image as high dimension vector or matrix, by projection or shift conversion be low-dimensional vector or matrix,
And vena metacarpea is indicated and matched under lower dimensional space.The method is sensitive to noise effects such as illumination variations, this limitation
Its application；(3) method based on textural characteristics mainly extracts image overall or local statistic information etc. as description,
Method versatility is stronger, efficiency high, and has good robustness for the displacement of small range, but for larger position
Displacement, its robust performance also has very big room for promotion；(4) method based on scale invariant feature, such method master
The widely used invariant features operator of computer vision field is come from, with certain robustness, but compares dependency graph picture
The preprocessing means such as enhancing, while computational efficiency is needed to be further improved.Four class methods respectively have advantage and disadvantage and advantage.
Method based on textural characteristics is that NBP algorithms are a kind of vena metacarpeas of use neighbour's binary pattern based on texture
Recognition methods.Vena metacarpea image is divided into some region units by this method, and calculation block average gray may during collection to eliminate
The image rotation influence of appearance, further realizes match cognization using NBP features.Specifically it may refer to document [1]：Lin Sen, Wu
Micro-, garden Wei fine jades studies [J] using the vena metacarpea living things feature recognition of texture neighbour's pattern.Chinese journal of scientific instrument, 2015,36
(10):2330-2338.Although the method has certain advantage in recognition speed, and is revolved in the displacement of small range and image
Turning upper has good robustness, but when displacement is larger, recognition effect is poor, and at this point between different people
Discrimination be not to know phenomenon by mistake it is obvious that easily causing, misclassification rate is high, is unfavorable for practical application.
Method based on scale invariant feature is that SIFT algorithms are that a kind of algorithm of computer vision is used for detecting and describing shadow
Locality characteristic as in, it finds extreme point in space scale, and extracts its position, yardstick, rotational invariants, this
Algorithm was delivered by David Lowe in 1999, is improved within 2004 and is summarized.Vena metacarpea identification side based on SIFT invariant features
Method is the extreme point for going out metric space using difference of Gaussian function check, and in the swollen characteristic point for choosing stabilization of these points, is
Each characteristic strip intends distribution direction, and then the feature generated is matched.This method used time is relatively more, in accuracy rate
Upper not special advantage, but effect is protruded in terms of different people palm vein is distinguished, and can effectively reduce misclassification rate.Cause
This, SIFT algorithms are effective supplements to NBP algorithms, how efficiently to combine both the above algorithm, while it is accurate to take into account identification
Rate and recognition time, and further reduction misclassification rate turns into concern emphatically in vena metacarpea Study of recognition to meet practical application
Research contents.
The content of the invention
It is an object of the invention to overcome the deficiencies in the prior art, propose that a kind of fusion textural characteristics and Scale invariant are special
The vena metacarpea recognition methods levied, while taking into account recognition speed (time), improves accuracy rate (the reduction knowledge by mistake of vena metacarpea identification
Rate).
For achieving the above object, the vena metacarpea recognition methods of present invention fusion textural characteristics and scale invariant feature,
It is characterised in that it includes following steps：
(1) (image enhaucament, is pre-processed to vena metacarpea ROI (the Region Of Interest) image in vena metacarpea storehouse
Processing), then extract NBP (textural characteristics) features of image and encoded, build NBP feature codings storehouse, meanwhile,
SIFT (scale invariant feature) feature of image is extracted, SIFT feature databases are built；
(2) NBP features, are extracted to vena metacarpea ROI image to be identified, and encoded；
(3), all NBP features in the NBP feature codings of vena metacarpea ROI image to be identified and NBP feature codings storehouse are compiled
Code is compared, and asks for Hamming distance, if smallest hamming distance is less than the threshold value t of setting1, show can obvious match cognization,
The corresponding vena metacarpea ROI image of smallest hamming distance NBP feature codings is then chosen as recognition result, terminates identification；It is no
Then, show can not obvious match cognization, then in NBP feature codings storehouse, choose h minimum vena metacarpea ROI of Hamming distance and scheme
As entering candidate regions；
(4) SIFT feature of vena metacarpea ROI image to be identified, and the h vena metacarpea ROI figures of candidate regions respectively, are extracted
As the SIFT feature in SIFT feature storehouse is matched, the i.e. SIFT Feature Points Matchings of matching result are obtained to quantity；
(5), SIFT feature and NBP characteristic bindings choose matching result：
When SIFT feature Point matching logarithm amount is more than or equal to threshold value t2, and SIFT feature Point matching logarithm amount maximum is
One, regard as in h vena metacarpea ROI image, matched with vena metacarpea ROI image to be identified, SIFT feature Point matching logarithm
Maximum vena metacarpea ROI image is measured as recognition result, when SIFT feature Point matching logarithm amount is more than or equal to threshold value t2, and
SIFT feature Point matching logarithm amount maximum more than one, then in h vena metacarpea ROI image, choose quiet with the palm to be identified
What Hamming distance was minimum between arteries and veins ROI image NBP feature codings is used as recognition result；
Take threshold value t2If SIFT feature matching is less than t to quantity2It is individual, then assert no identification object.
The object of the present invention is achieved like this.
The vena metacarpea recognition methods of present invention fusion textural characteristics and invariant features feature, is carried out just by NBP algorithms
Sieve, if primary dcreening operation fruiting area is clearly aobvious to have the threshold value t that smallest hamming distance is less than setting1, then corresponding vena metacarpea is chosen
ROI image as vena metacarpea ROI image to be identified recognition result；If primary dcreening operation result distinguishes not clear aobvious, special using SIFT
Carry out postsearch screening is levied, if aobvious i.e. SIFT feature Point matching logarithm amount is more than or equal to threshold value t clearly for postsearch screening fruiting area2,
Then it is identified result (identical if there is quantity, what Hamming distance was minimum in them are chosen is used as matching result)；When two
When secondary screening substantially can not distinguish result, then it is assumed that without identification object.The present invention can both lift exclusive use NBP and calculate
The accuracy rate that method and SIFT algorithms carry out vena metacarpea identification is to reduce misclassification rate, can meet collection in real time in efficiency again
The requirement of identification is to have taken into account recognition speed, so overcomes that NBP algorithm misclassification rates are higher and SIFT algorithms are time-consuming lacks
Point.
Brief description of the drawings
Fig. 1 is a kind of specific embodiment party of vena metacarpea recognition methods of present invention fusion textural characteristics and scale invariant feature
Formula flow chart；
Fig. 2 is the compares figure before and after the enhancing of vena metacarpea ROI image, wherein, (a) is the preceding i.e. artwork of enhancing, and (b) is enhancing
Afterwards；
Fig. 3 is that block divides schematic diagram；
Fig. 4 is that NBP encodes an instantiation schematic diagram；
Fig. 5 is SIFT feature matching figure；
Fig. 6 is existing based on RANSC algorithms, the method flow diagram rejected to wrong SIFT feature Point matching；
Fig. 7 is the flow chart for the method rejected to wrong SIFT feature Point matching in the present invention；
Fig. 8 is the comparison diagram of two kinds of proposition methods.
Embodiment
The embodiment to the present invention is described below in conjunction with the accompanying drawings, so that those skilled in the art is more preferable
Ground understands the present invention.Requiring particular attention is that, in the following description, when known function and the detailed description of design
When perhaps can desalinate the main contents of the present invention, these descriptions will be ignored herein.
Fig. 1 is a kind of specific embodiment party of vena metacarpea recognition methods of present invention fusion textural characteristics and scale invariant feature
Formula flow chart.
In the present embodiment, as shown in figure 1, the vena metacarpea identification of present invention fusion textural characteristics and scale invariant feature
Method comprises the following steps：
Step S1：Set up NBP feature codings storehouse and SIFT feature storehouse
NBP is set up to vena metacarpea ROI (Region Of Interest) image by pretreatment (image enhancement processing)
(textural characteristics) feature coding storehouse and SIFT (scale invariant feature) feature database.
In the matching process, in order to improve recognition speed and efficiency, we are first carried out in advance to the ROI image in vena metacarpea storehouse
Processing, i.e., carry out enhancing processing, as shown in Fig. 2 wherein, Fig. 2 (a) is vena metacarpea ROI image artwork, Fig. 2 (b) is to image
Vena metacarpea ROI image after image enhaucament.Comparison diagram 2 (a), Fig. 2 (b) it will be seen that image vena metacarpea is relatively sharp can
Distinguish.
After pretreatment, the NBP (textural characteristics) and SIFT (scale invariant feature) feature of image are extracted, NBP is built special
Levy code database and SIFT feature storehouse.
Step S2：NBP features are extracted to vena metacarpea ROI image to be identified, and encoded.
In step S1, step S2, it is encoded to extracting NBP features：
1.1) block division, is carried out to vena metacarpea ROI image
First, image block square formation V is converted into using the vena metacarpea ROI image of M × M size as matrix, i.e.,：
Wherein, each block Vij(i, j=1,2 ..., k) be m × m size square formation, M=k × m；
Then, the gray value of each block is calculated with formula (2)：
Wherein, fij(x, y) is block VijThe interior gray value as (x, y) place pixel, the gray value of all blocks can
To form the vena metacarpea image polylith Mean Matrix I of k × k sizes：
In the present embodiment, as shown in figure 3, the vena metacarpea ROI image of 128 × 128 sizes is converted as matrix
For the pixel average in 16 × 16 image block square formation V (now k=16, m=8), and difference calculation block, average is made
It is vena metacarpea image polylith Mean Matrix I to build 16 × 16 new matrix for pixel value.From figure 3, it can be seen that image is done
Polylith mean value computation, per small images image entirety and local message can be done simultaneously hold, reduce IMAQ when because of
The influence caused is rotated by a small margin, and substantially reduces original data volume, operation efficiency is improved, without the area of effect characteristicses
Indexing.
1.2) NBP encoding operations, are carried out to vena metacarpea image polylith Mean Matrix I, NBP binary-coding bits are formed
String；
NBP features are absorbed in the gray scale magnitude relationship between neighbor pixel.Described NBP encoding operations are, for vena metacarpea
Image polylith Mean Matrix I, using each element as a central element, takes the window of one 3 × 3, around central element,
Using top left hand element as starting point, using the method traveled through clockwise, 8 points around central element are extracted (if first
Row, first row, last column, last row, then window is not on vena metacarpea image polylith Mean Matrix I, then value is
0), it is in line side by side, element value is followed successively by p7,p6,…,p0；Then, since the element value of the leftmost side, by currentElement
Value is compared with next neighbour's element value on the right side of it, and formula is：
Especially,
The NBP codings of each central element are connected, as the NBP feature codings of vena metacarpea ROI image.
In the present embodiment, as shown in figure 4, to carrying out NBP codings at central element, NBP is obtained using formula (4) (5)
It is encoded to：00011010.16 × 16 NBP amounted at 256 elements are calculated successively to encode, and are arranged in order and are obtained 2048 NBP
Coding.
The NBP feature codings of all vena metacarpea ROI images in vena metacarpea storehouse are connected, you can composition NBP features
Code database.
Step S3：The NBP feature codings of vena metacarpea ROI image to be identified are special with all NBP in NBP feature codings storehouse
Assemble-publish code is compared, and asks for Hamming distance.
For the NBP feature codings of two comparisons, S is designated asNBP1, SNBP2, its Bit String form is：
SNBP1=a1a2…aN (6)
SNBP2=b1b2…bN (7)
Wherein, a1~aN, b1~bNFor 0 or 1；
Hamming distance between them is defined as：
Wherein, symbolXOR is represented, N is the length of NBP feature codings.
If smallest hamming distance is less than the threshold value t of setting1, show substantially recognize matching, then choose the minimum Hamming
Apart from the corresponding vena metacarpea ROI image of NBP feature codings as matching result, terminate identification；Otherwise, show substantially to know
Do not match, then in NBP feature codings storehouse, choose h minimum vena metacarpea ROI image of Hamming distance and enter candidate regions.
In the present embodiment, Hamming distance is asked for, if smallest hamming distance is less than or equal to threshold value t1=0.21, then choose
Correspondence image is recognition result in vena metacarpea storehouse, and terminates identification, if smallest hamming distance is more than t1=0.21, then choose the palm
The h=20 vena metacarpea ROI image minimum with vena metacarpea ROI image Hamming distance to be identified enters candidate regions in vein storehouse.
In the present embodiment, as shown in figure 5, wherein, (a) is vena metacarpea ROI image to be identified, (b), (c) are when carrying out primary dcreening operation
Two representative width in the 20 width vena metacarpea ROI images selected.
Step S4：Extract the SIFT feature of vena metacarpea ROI image to be identified, and h vena metacarpea of candidate regions respectively
SIFT feature of the ROI image in SIFT feature storehouse is matched, and obtains the i.e. SIFT Feature Points Matchings of matching result to quantity.
In the present embodiment, in step S1, step S4, the SIFT feature of the extraction vena metacarpea ROI image is：
2.1) difference of Gaussian pyramid, is built, extreme point is found
In order to find the characteristic point unrelated with yardstick, first have to construct image pyramid.Image pyramid passes through to original
Image continuously does 1/2nd sampling constructions and formed.The difference of Gaussian for calculating the pyramid diagram picture of adjacent yardstick again obtains Gauss
Difference image, so obtains that extreme point can be found in the image space after a series of difference of Gaussian images.Wherein, Gauss
Difference pyramid constructive formula is：
D (x, y, σ)=(G (x, y, k σ)-G (x, y, σ)) * I (x, y)
=L (x, y, k σ)-L (x, y, σ) (9)
Wherein, D (x, y, σ) is changeable scale Gaussian function, and G (x, y, σ) is Gaussian convolution core, and L (x, y, σ) is image
Metric space, (x, y) is space coordinate, and σ sizes determine the smoothness of image, the general picture feature of large scale correspondence image,
The minutia of small yardstick correspondence image.
Find extreme point method be：Test point is corresponding with 8 consecutive points of yardstick and neighbouring yardstick with it
Totally 26 points compare 18 points, and it is extreme point that it is determined that when it is all bigger or all small than other adjacent domains points.
2.2), positioning feature point is with removing skirt response
2.2.1), characteristic point is accurately positioned
Because DoG values are to noise-sensitive, therefore detect Local Extremum in DoG metric spaces and also to pass through into one
Step, which is examined, can just be defined as characteristic point.DoG functions are in the Taylor expansions of metric space：
To above formula derivation, and make it be 0, obtain accurate position, obtain：
Formula (11) is substituted into formula (10), formula is obtained：
WhenDuring less than q1, cast out this feature point.Q1 advises taking 0.03 herein.
2.2.2 skirt response), is removed
Hessian matrixes are built,
WhenWhen, characteristic point is retained, it is on the contrary then rejected.R advises that value is 10 herein.
2.3), direction is distributed
SIFT is characterized an imparting direction by the gradient of each characteristic point.The definition of gradient magnitude is：
The definition of gradient direction is：
Determine that the direction of characteristic point uses histogram of gradients statistic law, that is, count in using characteristic point as origin certain area
Image slices vegetarian refreshments made contribution is generated to characteristic point direction.Histogram of gradients is using every s degree direction as a post, common 360/s
Individual post, the direction representated by post is pixel gradient direction, and the length of post represents gradient magnitude.The maximum side of gradient magnitude
To the direction of as this feature point.
2.4), generation feature point description
The feature point description period of the day from 11 p.m. to 1 a.m is generated, reference axis is first rotated to the direction of characteristic point, to ensure rotational invariance.So
Afterwards to arbitrary characteristics point, the region of 16 × 16 sizes centered on characteristic point is taken in its metric space, each 4 × 4 it is small
The gradient vector histogram in 8 directions is calculated on block, the accumulated value of each gradient direction is drawn, a seed point region is formed.
One characteristic point using 4 × 4 totally 16 points describe, form the characteristic vector of one 128 dimension.Again by the length of characteristic vector
Normalization, further removes the influence of illumination variation, generates final SIFT feature description.
In the 20 width vena metacarpea ROI images that Fig. 5 (d), (e), (f) be respectively vena metacarpea ROI image to be identified, is selected
Two representative width SIFT feature description figure be (a), (b), the SIFT feature point description figures of (c), feature points
Mesh is respectively 520,491,576.SIFT feature is asked for each image, difference of Gaussian pyramid is built first, then to Gauss
Image asks for extreme point, and characteristic point is accurately positioned and removed after skirt response, continues as characteristic point and assigns direction, generates feature
Description.The characteristic point of each image such as Fig. 5 (d), (e), (f) are shown, have 500 or so characteristic points, data volume is larger,
Each matching of SIFT feature all needs the matching two-by-two between characteristic point, therefore workload can be greatly reduced by carrying out primary dcreening operation.
When carrying out characteristic matching, matched using closest range search method, that is, find each characteristic point
2 arest neighbors characteristic points, if minimum distance is less than some proportion threshold value t with secondary ratio closely3, then receive this pair
Match point.So, matching result i.e. SIFT feature Point matching logarithm amount is obtained.
After SIFT matchings are carried out to image, some unreliable SIFT feature Point matchings pair can have been obtained unavoidably.When unreliable
Feature Points Matching when rising to quantity, images match accuracy will be caused to reduce or failure.So, in SIFT feature
With rear, it is necessary to SIFT feature Point matching to screening, reject the SIFT feature Point matching pair of mistake, lifting images match
Accuracy.
The elimination method of traditional wrong SIFT feature Point matching pair uses RANSC algorithms.RANSAC algorithms are based on
Two it is assumed that one is that data set includes interior point (inliers) and exterior point (outliers)；Two be that exterior point does not meet and estimated by interior point
The model of meter.It randomly selects a computational mathematics model first, then finds the interior point of the model, recalculates model, pass through
Constantly iterate to calculate to find a model for meeting most multiple spot.
If image I1And I2SIFT feature Point matching to collection (abbreviation matching is to integrating) as φ.Smallest sample collection S is to count
The minimum pairing set of all parameters of model is calculated, because the transformation model of image there are 8 frees degree, so needing at least four special
Point matching is levied to including 4 Feature Points Matchings pair in transformation matrix, therefore S to calculate.It is less than certain threshold with the residual error of matrix H
The collection of the point composition of value is collectively referred to as the consistent collection (Consensus Set) of matrix H, is set to CS.
RANSAC algorithm flow charts such as Fig. 6, it is concretely comprised the following steps：
(1) from initial matching to randomly selecting a smallest sample collection S in collection φ；
(2) homography matrix H is calculated according to S；
(3) whether each point meets model M in φ in being calculated according to error metrics function, and the point set composition H's met is consistent
Collect CS, and return to matching logarithm n in CS；
(4) according to CS again homography matrix H, step (3) calculating, the number of repetition k given until reaching are then back to；
(5) take in k calculating, consistent collection maximum n is used as final interior point data collection CSm；
(6) CS is usedmHomography matrix H is calculated, the data model that set φ is final is used as.
This method can effectively exclude error hiding pair, but can abandon some correct matchings pair, and in mistake
Match point it is more or matching logarithm it is less when all can cause some troubles to final matching result.
In the present embodiment, the mode of RANSAC algorithms is used for reference, a kind of SIFT feature for vena metacarpea image is proposed
Match to screening mode.Before vena metacarpea identification, its ROI image is that, by normalized, will not occur big
Displacement and rotation, utilize the two characteristic point lines and the angle of horizontal direction and two features of SIFT feature Point matching pair
The distance between point screens the SIFT feature Point matching pair made mistake.We define two straight line L1、L2Similarity distance
For：
g(L1,L2)=γ1l(L1,L2)+γ2θ(L1,L2)
Wherein, l (L1,L2) it is two straight line L1、L2The difference of length, θ (L1,L2) it is two straight line L1、L2Between folder
Angle, γ1For apart from weights, γ2For angle weights, apart from weights γ1With the product and angle weights γ of difference2With angle
Product is dimensionless number.That is, two straight line L1、L2The similarity distance weighted sum that both is exactly.
In the present embodiment, to wrong SIFT feature Point matching to the flow chart of method rejected as shown in Fig. 7,
Specific implementation step is：
3.1), all SIFT feature Point matchings are ranked up to constituting a pairing set φ, with SIFT feature Point matching pair
2 lines and the angle of horizontal direction be the first priority, be ranked up, choose by the second priority of distance between two points
Preceding n groups SIFT feature Point matching is to being used as a sample set S；
3.2) 2 average distances and average angle of SIFT feature Point matching pair in sample set S, are calculated, with average value
It is used as straight line L1；
3.3) 2 line L of SIFT feature Point matching pair in a pairing set φ, are calculated2With straight line L1Similarity away from
From：
g(L1,L2)=γ1l(L1,L2)+γ2θ(L1,L2)
Wherein, l (L1,L2) it is two straight line L1、L2The difference of length, θ (L1,L2) it is two straight line L1、L2Between folder
Angle, γ1For apart from weights, γ2For angle weights, apart from weights γ1With the product and angle weights γ of difference2With angle
Product is dimensionless；
Given threshold z, calculates the SIFT feature Point matching comparative example k that similarity distance is less than threshold value z；
If 3.4), k is more than or equal to the threshold value k of setting1, then eligible i.e. similarity distance is less than all of threshold value z
SIFT feature Point matching is to as the result after rejecting, terminating to reject；Otherwise step 3.5 is entered)；
If 3.5), k is more than or equal to the threshold value k of setting2, then choose and eligible i.e. similarity distance be less than threshold value z's
All SIFT feature Point matchings are used as benchmark, return to step 3.2 to sample set S), otherwise into step 3.6)；Wherein, threshold value
Threshold value k1More than threshold value k2；
3.6) SIFT characteristic points maximum to similarity distance with other SIFT feature Point matchings in sample set S, are removed
Pairing, the next SIFT feature Point matching pair added after being sorted in a pairing set φ, return to step 3.2), if executed
Last SIFT feature Point matching pair after being sorted into a pairing set φ, without next SIFT feature Point matching pair, then
Choose in iterative process, ratio k highest sample set S terminate to reject as the result after rejecting.
In the present embodiment, as shown in figure 8, the rejecting result to two methods is compared.In the present embodiment, select
Group number n is 4, threshold value k1For 90%, threshold value k2Rejected for 30%.Fig. 8 (a) represents the original matching knot without rejecting
Really, Fig. 8 (b) is the result after RANSAC methods are rejected, and Fig. 8 (c) is the result after method rejecting in the present embodiment.It can see
Arrive, the first row, when there is more SIFT feature Point matching pair, RANSAC can abandon some correct SIFT feature Point matchings pair,
And the elimination method that the present embodiment is used, can reject wrong SIFT feature Point matching to while retain relatively large number of
Correct SIFT Feature Points Matchings pair；Second row, in erroneous matching, RANSAC methods can find the H-matrix of partial error, protect
The elimination method for staying a small amount of erroneous matching the present embodiment to use can reject whole erroneous matchings；The third line, in SIFT feature
When matching is relatively fewer to quantity, the result of RANSAC methods retains less, and criterion of identification can not be reached when threshold value is higher,
And the elimination method that the present embodiment is used can retain more matching, vena metacarpea can be correctly recognized.
(5), SIFT feature and NBP characteristic bindings choose matching result：
When SIFT feature Point matching logarithm amount is more than or equal to threshold value t2, and SIFT feature Point matching logarithm amount maximum is
One, regard as in h vena metacarpea ROI image, matched with vena metacarpea ROI image to be identified, SIFT feature Point matching logarithm
Maximum vena metacarpea ROI image is measured as matching result, when SIFT feature Point matching logarithm amount is more than or equal to threshold value t2, and
SIFT feature Point matching logarithm amount maximum more than one, then in h vena metacarpea ROI image, choose quiet with the palm to be identified
What Hamming distance was minimum between arteries and veins ROI image NBP feature codings is used as matching result；
Take threshold value t2If SIFT feature matching is less than t to quantity2It is individual, then assert without matching object.
In vena metacarpea identification process, as shown in Figure 5.Fig. 5 (a) and the vena metacarpea ROI image that (b) is same people, both
Between have larger displacement, during NBP primary dcreening operations, Fig. 5 (a) and Fig. 5 (b) Hamming distance are 0.2505, Fig. 5 (a) and figure
The Hamming distance of 5 (c) is 0.2485, only using only NBP algorithms, it will Fig. 5 (a) is mistakenly considered Fig. 5 (c) and obtains wrong conclusion.
And using the recognition methods of the present invention, although both Hamming distances are big after the above two Hamming distance ratios, but still will in primary dcreening operation
(b) (c), which is picked out, enters candidate regions, when carrying out SIFT feature matching to both, and such as Fig. 5 (g) and Fig. 5 (h) are shown, preceding
Both SIFT matching characteristics points are to for 83 pairs, then both matching characteristic points are to for 3 pairs, so choosing correct matching figure
As being Fig. 5 (a) and Fig. 5 (b), that is, it is identified as Fig. 5 (b).
The present invention has apparent advantage, not only to former SIFT algorithms in vena metacarpea identification on recognition time
The image for substantially distinguishing NBP algorithms by primary dcreening operation is rejected, more by the image SIFT feature matching amount that can not substantially distinguish by original
500 times come are reduced to 20 times, so as to greatly reduce match time.For NBP algorithms, this algorithm can be to former algorithm
Area is distributed unconspicuous image and carries out postsearch screening, is even more to improve bright in terms of reducing the error rate of NBP algorithms, erroneous matching
It is aobvious.In actual applications, the inventive method has apparent advantage.
| Method | Average time | Misclassification rate (1：N) | Reject rate (1:1) |
| SIFT algorithms | 31.7s | 0.02% | 5.0% |
| NBP algorithms | 6.6ms | 1.8% | 1.6% |
| The present invention | 0.1s | 0% | 0.6% |
Table 1
Table 1 is the present invention with merging data comparison form of the front method on Hong Kong Chinese University's vena metacarpea storehouse.From table 1
In as can be seen that the present invention compared to NBP algorithms in identification process, recognition performance can be effectively improved, especially in mistake
In terms of identification, the present invention can reject the characteristic of erroneous matching very well using SIFT algorithms, effectively reduce erroneous matching；This
Method compares SIFT algorithms, big advantage is occupied on match time, this method match time is less, disclosure satisfy that reality
The requirement of match time in.Therefore, SIFT algorithms and NBP algorithms are compared, the present invention is capable of two kinds of algorithms of effective integration
Advantage, all has some improvement on matching speed and conventional efficient.
Table 2 adopts vena metacarpea storehouse certainly
| Method | Average time | Misclassification rate (1：N) | Reject rate (1:1) |
| SIFT algorithms | 29.2s | 0.05% | 4.5% |
| NBP algorithms | 6.6ms | 2.25% | 3.0% |
| The present invention | 0.09s | 0% | 1.5% |
Table 2
Table 2 be the present invention with merge in front of method from the data comparison form adopted on vena metacarpea storehouse, can from table 2
Arrive, in storehouse is adopted certainly, preferable improvement can equally be obtained using the present invention.
Although illustrative embodiment of the invention is described above, in order to the skill of the art
Art personnel understand the present invention, it should be apparent that the invention is not restricted to the scope of embodiment, to the general of the art
For logical technical staff, as long as various change is in the spirit and scope of the present invention that appended claim is limited and is determined,
These changes are it will be apparent that all utilize the innovation and creation of present inventive concept in the row of protection.
Claims (3)
1. a kind of vena metacarpea recognition methods for merging textural characteristics and scale invariant feature, it is characterised in that comprise the following steps：
(1), vena metacarpea ROI (the Region Of Interest) image in vena metacarpea storehouse is pre-processed (at image enhaucament
Reason), then extract NBP (textural characteristics) features of image and encoded, build NBP feature codings storehouse, meanwhile, extract
SIFT (scale invariant feature) feature of image, builds SIFT feature storehouse；
(2) NBP features, are extracted to vena metacarpea ROI image to be identified, and encoded；
(3), the NBP feature codings of vena metacarpea ROI image to be identified and all NBP feature codings in NBP feature codings storehouse enter
Row is compared, and asks for Hamming distance, if smallest hamming distance is less than the threshold value t of setting1, show can obvious match cognization, then select
Take the corresponding vena metacarpea ROI image of smallest hamming distance NBP feature codings as recognition result, terminate identification；Otherwise, show
Can not obvious match cognization, then in NBP feature codings storehouse, choose h minimum vena metacarpea ROI image of Hamming distance and enter time
Constituency；
(4) SIFT feature of vena metacarpea ROI image to be identified, is extracted, and h vena metacarpea ROI image of candidate regions exists respectively
SIFT feature in SIFT feature storehouse is matched, and obtains matching result i.e. SIFT feature Point matching logarithm amount；
(5), SIFT feature and NBP characteristic bindings choose matching result：
When SIFT feature Point matching logarithm amount is more than or equal to threshold value t2, and SIFT feature Point matching logarithm amount maximum is one, is recognized
It is set in h vena metacarpea ROI image, is matched with vena metacarpea ROI image to be identified, SIFT feature Point matching logarithm amount maximum
Vena metacarpea ROI image is as recognition result, when SIFT feature Point matching logarithm amount is more than or equal to threshold value t2, and SIFT feature
Quantity maximum more than one is matched, then in h vena metacarpea ROI image, is chosen and vena metacarpea ROI image NBP to be identified
What Hamming distance was minimum between feature coding is used as recognition result；
Take threshold value t2If SIFT feature matching is less than t to quantity2It is individual, then assert no identification object.
2. vena metacarpea recognition methods according to claim 1, it is characterised in that, it is necessary to wrong SIFT in step (4)
Feature Points Matching is concretely comprised the following steps to rejecting：
3.1), to all SIFT feature Point matchings to constitutingPairing set φIt is ranked up, with the two of SIFT feature Point matching pair
The angle for putting line and horizontal direction is the first priority, is ranked up by the second priority of distance between two points, n groups before choosing
SIFT feature Point matching is to being used as a sample set S；
3.2) 2 average distances and average angle of SIFT feature Point matching pair in sample set S, are calculated, using average value as straight
Line L1；
3.3) 2 line L of SIFT feature Point matching pair in a pairing set φ, are calculated2With straight line L1Similarity distance：
g(L1,L2)=γ1l(L1,L2)+γ2θ(L1,L2)
Wherein, l (L1,L2) it is two straight line L1、L2The difference of length, θ (L1,L2) it is two straight line L1、L2Between angle, γ1
For apart from weights, γ2For angle weights, apart from weights γ1With the product and angle weights γ of difference2Product with angle is equal
For dimensionless；
Given threshold z, calculates the SIFT feature Point matching comparative example k that similarity distance is less than threshold value z；
If 3.4), k is more than or equal to the threshold value k of setting1, then eligible i.e. similarity distance is less than to threshold value z all SIFT spies
Point matching is levied to as the result after rejecting, terminating to reject；Otherwise step 3.5 is entered)；
If 3.5), k is more than or equal to the threshold value k of setting2, then choose and eligible i.e. similarity distance be less than all of threshold value z
SIFT feature Point matching is used as benchmark, return to step 3.2 to sample set S), otherwise into step 3.6)；Wherein, threshold k1
More than threshold value k2；
3.6) SIFT feature Point matching pair maximum to similarity distance with other SIFT feature Point matchings in sample set S, is removed,
Add a pairing set φ in sort after next SIFT feature Point matching pair, return to step 3.2), if executed to match
Last SIFT feature Point matching pair after to being sorted in collection φ, without next SIFT feature Point matching pair, then chooses and changes
During generation, ratio k highest sample set S terminate to reject as the result after rejecting.
3. vena metacarpea recognition methods according to claim 1, described group of number n is 4, threshold value k1For 90%, threshold value k2For
30%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710222874.5A CN107145829B (en) | 2017-04-07 | 2017-04-07 | Palm vein identification method integrating textural features and scale invariant features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710222874.5A CN107145829B (en) | 2017-04-07 | 2017-04-07 | Palm vein identification method integrating textural features and scale invariant features |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107145829A true CN107145829A (en) | 2017-09-08 |
| CN107145829B CN107145829B (en) | 2020-05-22 |
Family
ID=59773689
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710222874.5A Expired - Fee Related CN107145829B (en) | 2017-04-07 | 2017-04-07 | Palm vein identification method integrating textural features and scale invariant features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107145829B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107980140A (en) * | 2017-10-16 | 2018-05-01 | 厦门中控智慧信息技术有限公司 | A kind of recognition methods of vena metacarpea and device |
| CN108710690A (en) * | 2018-05-22 | 2018-10-26 | 长春师范大学 | Medical image search method based on geometric verification |
| CN109961078A (en) * | 2017-12-22 | 2019-07-02 | 展讯通信（上海）有限公司 | Images match and joining method, device, system, readable medium |
| CN110049097A (en) * | 2019-03-04 | 2019-07-23 | 平安科技（深圳）有限公司 | Information-pushing method, device, server and computer storage medium |
| CN110348289A (en) * | 2019-05-27 | 2019-10-18 | 广州中国科学院先进技术研究所 | A kind of finger vein identification method based on binary map |
| CN110705341A (en) * | 2019-08-13 | 2020-01-17 | 平安科技（深圳）有限公司 | Verification method, device and storage medium based on finger vein image |
| CN110717469A (en) * | 2019-10-16 | 2020-01-21 | 山东浪潮人工智能研究院有限公司 | Finger vein identification method and system based on correlation semantic feature learning |
| CN112200159A (en) * | 2020-12-01 | 2021-01-08 | 四川圣点世纪科技有限公司 | Non-contact palm vein identification method based on improved residual error network |
| CN114998950A (en) * | 2022-08-01 | 2022-09-02 | 北京圣点云信息技术有限公司 | Vein encryption and identification method based on deep learning |
| CN115311691A (en) * | 2022-10-12 | 2022-11-08 | 山东圣点世纪科技有限公司 | Joint identification method based on wrist vein and wrist texture |
| CN115631514A (en) * | 2022-10-12 | 2023-01-20 | 中海银河科技(北京)有限公司 | Palm vein fingerprint-based user identification method, device, equipment and medium |
| TWI794132B (en) * | 2022-09-19 | 2023-02-21 | 威盛電子股份有限公司 | System for detecting misidentified objects |
Citations (8)
| 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 |
| CN102622587A (en) * | 2012-03-08 | 2012-08-01 | 哈尔滨工程大学 | Hand back vein recognition method based on multi-scale second-order differential structure model and improved watershed algorithm |
| US20120288166A1 (en) * | 2011-05-13 | 2012-11-15 | Microsoft Corporation | Association and prediction in facial recognition |
| CN103136751A (en) * | 2013-02-05 | 2013-06-05 | 电子科技大学 | Improved scale invariant feature transform (SIFT) image feature matching algorithm |
| CN103729654A (en) * | 2014-01-22 | 2014-04-16 | 青岛新比特电子科技有限公司 | Image matching retrieval system on account of improving Scale Invariant Feature Transform (SIFT) algorithm |
| CN104616035A (en) * | 2015-03-12 | 2015-05-13 | 哈尔滨工业大学 | Visual Map rapid matching method based on global image feature and SURF algorithm |
| CN105117712A (en) * | 2015-09-15 | 2015-12-02 | 北京天创征腾信息科技有限公司 | Single-sample human face recognition method compatible for human face aging recognition |
| CN105760815A (en) * | 2016-01-26 | 2016-07-13 | 南京大学 | Heterogeneous human face verification method based on portrait on second-generation identity card and video portrait |
-
2017
- 2017-04-07 CN CN201710222874.5A patent/CN107145829B/en not_active Expired - Fee Related
Patent Citations (8)
| 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 |
| US20120288166A1 (en) * | 2011-05-13 | 2012-11-15 | Microsoft Corporation | Association and prediction in facial recognition |
| CN102622587A (en) * | 2012-03-08 | 2012-08-01 | 哈尔滨工程大学 | Hand back vein recognition method based on multi-scale second-order differential structure model and improved watershed algorithm |
| CN103136751A (en) * | 2013-02-05 | 2013-06-05 | 电子科技大学 | Improved scale invariant feature transform (SIFT) image feature matching algorithm |
| CN103729654A (en) * | 2014-01-22 | 2014-04-16 | 青岛新比特电子科技有限公司 | Image matching retrieval system on account of improving Scale Invariant Feature Transform (SIFT) algorithm |
| CN104616035A (en) * | 2015-03-12 | 2015-05-13 | 哈尔滨工业大学 | Visual Map rapid matching method based on global image feature and SURF algorithm |
| CN105117712A (en) * | 2015-09-15 | 2015-12-02 | 北京天创征腾信息科技有限公司 | Single-sample human face recognition method compatible for human face aging recognition |
| CN105760815A (en) * | 2016-01-26 | 2016-07-13 | 南京大学 | Heterogeneous human face verification method based on portrait on second-generation identity card and video portrait |
Non-Patent Citations (5)
| Title |
|---|
| MEIRU MU 等: "Sift and gray scale invariant features for palmprint indentification using complex directional wavelet and local binary pattern", 《NEUROCOMPUTING》 * |
| PIERRE-OLIVIER LADOUX 等: "Plam vein verification system based on SIFT matching", 《INTERNATIONAL CONFERENCE ON BIOMETRICS:ICB2009:ADVANCES IN BIOMETRICS》 * |
| 张德全 等: "改进SIFT算法在图像数据相似性匹配中的应用研究", 《计算机科学》 * |
| 林森 等: "采用纹理近邻模式的掌静脉生物特征识别研究", 《仪器仪表学报》 * |
| 蓝惠英: "基于LavVIEW的手掌静脉身份识别系统研究", 《中国优秀硕士学位论文全文数据库-信息科技辑》 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107980140B (en) * | 2017-10-16 | 2021-09-14 | 厦门熵基科技有限公司 | Palm vein identification method and device |
| CN107980140A (en) * | 2017-10-16 | 2018-05-01 | 厦门中控智慧信息技术有限公司 | A kind of recognition methods of vena metacarpea and device |
| CN109961078A (en) * | 2017-12-22 | 2019-07-02 | 展讯通信（上海）有限公司 | Images match and joining method, device, system, readable medium |
| CN110490271A (en) * | 2017-12-22 | 2019-11-22 | 展讯通信（上海）有限公司 | Images match and joining method, device, system, readable medium |
| CN110490271B (en) * | 2017-12-22 | 2021-09-21 | 展讯通信（上海）有限公司 | Image matching and splicing method, device, system and readable medium |
| CN108710690A (en) * | 2018-05-22 | 2018-10-26 | 长春师范大学 | Medical image search method based on geometric verification |
| CN110049097A (en) * | 2019-03-04 | 2019-07-23 | 平安科技（深圳）有限公司 | Information-pushing method, device, server and computer storage medium |
| CN110348289A (en) * | 2019-05-27 | 2019-10-18 | 广州中国科学院先进技术研究所 | A kind of finger vein identification method based on binary map |
| CN110705341A (en) * | 2019-08-13 | 2020-01-17 | 平安科技（深圳）有限公司 | Verification method, device and storage medium based on finger vein image |
| WO2021027155A1 (en) * | 2019-08-13 | 2021-02-18 | 平安科技（深圳）有限公司 | Verification method and apparatus based on finger vein image, and storage medium and computer device |
| CN110717469B (en) * | 2019-10-16 | 2022-04-12 | 山东浪潮科学研究院有限公司 | Finger vein identification method and system based on correlation semantic feature learning |
| CN110717469A (en) * | 2019-10-16 | 2020-01-21 | 山东浪潮人工智能研究院有限公司 | Finger vein identification method and system based on correlation semantic feature learning |
| CN112200159A (en) * | 2020-12-01 | 2021-01-08 | 四川圣点世纪科技有限公司 | Non-contact palm vein identification method based on improved residual error network |
| CN114998950A (en) * | 2022-08-01 | 2022-09-02 | 北京圣点云信息技术有限公司 | Vein encryption and identification method based on deep learning |
| TWI794132B (en) * | 2022-09-19 | 2023-02-21 | 威盛電子股份有限公司 | System for detecting misidentified objects |
| CN115311691A (en) * | 2022-10-12 | 2022-11-08 | 山东圣点世纪科技有限公司 | Joint identification method based on wrist vein and wrist texture |
| CN115631514A (en) * | 2022-10-12 | 2023-01-20 | 中海银河科技(北京)有限公司 | Palm vein fingerprint-based user identification method, device, equipment and medium |
| CN115311691B (en) * | 2022-10-12 | 2023-02-28 | 山东圣点世纪科技有限公司 | Joint identification method based on wrist vein and wrist texture |
| CN115631514B (en) * | 2022-10-12 | 2023-09-12 | 中海银河科技(北京)有限公司 | User identification method, device, equipment and medium based on palm vein fingerprint |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107145829B (en) | 2020-05-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107145829A (en) | A kind of vena metacarpea recognition methods for merging textural characteristics and scale invariant feature | |
| CN104915949B (en) | A kind of image matching method of combination point feature and line feature | |
| CN105931252B (en) | A kind of oval rapid detection method based on geometrical constraint | |
| CN106355577B (en) | Rapid image matching method and system based on significant condition and global coherency | |
| Reddy Mopuri et al. | Object level deep feature pooling for compact image representation | |
| CN107122712B (en) | Palm print image identification method based on CNN and bidirectional VLAD | |
| CN109118528A (en) | Singular value decomposition image matching algorithm based on area dividing | |
| CN109299720A (en) | A kind of target identification method based on profile segment spatial relationship | |
| CN107341813A (en) | SAR image segmentation method based on structure learning and sketch characteristic inference network | |
| CN107392215A (en) | A kind of multigraph detection method based on SIFT algorithms | |
| CN110321769A (en) | A kind of more size commodity on shelf detection methods | |
| Roy et al. | Script identification from handwritten document | |
| CN107016680B (en) | A kind of pest image background minimizing technology based on conspicuousness detection | |
| CN106650580A (en) | Image processing based goods shelf quick counting method | |
| CN104361096A (en) | Image retrieval method based on characteristic enrichment area set | |
| CN109325510B (en) | Image feature point matching method based on grid statistics | |
| CN110866439A (en) | Hyperspectral image joint classification method based on multi-feature learning and superpixel kernel sparse representation | |
| CN105069459B (en) | One kind is directed to High Resolution SAR Images type of ground objects extracting method | |
| CN114511012A (en) | SAR image and optical image matching method based on feature matching and position matching | |
| CN115731257A (en) | Leaf form information extraction method based on image | |
| CN114387592A (en) | Character positioning and identifying method under complex background | |
| CN105631860B (en) | Image point extracting method of the same name based on partial ordering's direction histogram description | |
| CN107292268A (en) | The SAR image semantic segmentation method of quick ridge ripple deconvolution Structure learning model | |
| CN107358244B (en) | A kind of quick local invariant feature extracts and description method | |
| CN111292346B (en) | Method for detecting contour of casting box body in noise environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200522 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |