CN112487867B - Visual constraint fingerprint identification method based on enhanced triangulation - Google Patents
Visual constraint fingerprint identification method based on enhanced triangulation Download PDFInfo
- Publication number
- CN112487867B CN112487867B CN202011209604.9A CN202011209604A CN112487867B CN 112487867 B CN112487867 B CN 112487867B CN 202011209604 A CN202011209604 A CN 202011209604A CN 112487867 B CN112487867 B CN 112487867B
- Authority
- CN
- China
- Prior art keywords
- minutiae
- matching
- triangle
- matching point
- fingerprint image
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000000007 visual effect Effects 0.000 title claims abstract description 26
- 239000013598 vector Substances 0.000 claims abstract description 27
- 238000012795 verification Methods 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 3
- 238000007670 refining Methods 0.000 claims 1
- 238000000605 extraction Methods 0.000 abstract description 2
- 238000013519 translation Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
- G06V40/1365—Matching; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
Abstract
The invention discloses a visual constraint fingerprint identification method based on enhanced triangulation, which comprises the steps of firstly preprocessing an input fingerprint picture, then carrying out feature extraction and triangulation to obtain a triangle set, and traversing domain minutiae determined by triangulation to form a triangle; removing repeated triangles to obtain a final triangle set; calculating the feature vector of the triangle; judging according to the feature vector to obtain a rough matching point set; using a double verification method to enable the matching points to correspond one by one; constructing adjacent local feature vectors to further verify accuracy and obtain a secondary matching point set; combining the visual characteristics, and removing the matching points again to obtain a final corresponding point set; and calculating a matching score, and judging whether the input fingerprint image and the template fingerprint image are successfully matched. The method considers the complex situations of losing, moving and the like of the minutiae, and combines with a method for eliminating mismatching points, thereby effectively improving the performance of fingerprint identification.
Description
Technical Field
The invention belongs to the technical field of biological feature identification and information security, and particularly relates to a visual constraint fingerprint identification method based on enhanced triangulation.
Background
Fingerprint recognition is one of important research contents of biological feature recognition. In recent years, fingerprint identification is widely used in actual life due to characteristics of lifelong invariance, accuracy, uniqueness and the like, such as a fingerprint access control system, a fingerprint attendance system and a mobile phone fingerprint identification system. However, in practical situations, the problems of scaling, translation, rotation and the like of the fingerprint image occur due to uneven force and different positions of the finger contacting the fingerprint acquisition instrument when the fingerprint is acquired. These can significantly impact the performance of the fingerprint identification method.
The fingerprint identification is to compare the input fingerprint with the fingerprints in the database to determine. The fingerprint identification mainly comprises the following steps of fingerprint acquisition, filtering, binarization, refinement, feature extraction, feature matching and the like.
The fingerprint recognition method based on the point mode generally calculates euclidean distance between feature vectors of minutiae points to obtain similarity between minutiae points, thereby judging whether two fingerprint images are matched. In practical situations, scaling, translation and rotation phenomena exist in the fingerprint image, so that problems such as omission, position deviation and the like exist in the extracted minutiae, and even false minutiae can be extracted. But the feature information between any minutiae in the fingerprint and neighboring minutiae is unchanged.
The fingerprint identification method based on enhanced triangulation (Expanded Delaunay Triangulation, EDT) improves the triangulation method. Firstly, triangulating an input detail point set to quickly obtain a triangle set; secondly, taking one minutiae as an example, obtaining all triangles formed by the minutiae; removing all the minutiae points that make up the triangles to form a new minutiae point set; and then, triangulating the newly obtained minutiae set again to obtain a triangle set. Performing the operation on all the detail points in the input detail point set to obtain triangle sets and solving the union of the triangle sets; finally, the identification is performed using a conventional point pattern matching method. (Mohamed Hedi Ghaddab, khaled Jouini, olajdi Korbaa. Fast and Accurate Fingerprint Matching Using Expanded Delaunay Triangulation [ C ]// IEEE/ACS International Conference on Computer Systems & applications. IEEE Computer Society, 2017.). Although triangulation takes triangle sets very fast, triangulation takes fewer triangle sets and contains fewer messages. Merely using EDT for fingerprinting may result in the presence of a mismatching minutiae point pair. The invention discovers the problem, proposes a visual constraint fingerprint identification method based on enhanced triangulation, improves EDT-C, and adds a method combining double matching and visual constraint to remove mismatching point pairs.
Disclosure of Invention
The invention aims to solve the problems of translation, rotation, nonlinear deformation and the like of a fingerprint image, and provides a visual constraint fingerprint identification method based on enhanced triangulation so as to improve the performance of the fingerprint identification method. According to the invention, the extracted detail point set is subjected to EDT triangulation to obtain a triangle set. Taking one minutiae point as an example, taking all triangles containing the minutiae point from a triangle set and forming other minutiae points forming the triangles into a set; the triangle is formed again by the minutiae and any other two minutiae in the composition set. Since two triangle sets are extracted, there are repeated triangles in the two triangle sets, the repeated triangles are removed from one of the triangle sets, and the two triangle sets are combined to form a final triangle set. Obtaining a rough matching point set according to the similarity of the triangle characteristics; in order to enable one-to-one matching of the input minutiae and template minutiae, a double matching method is used to remove one-to-many or many-to-one phenomena. In order to avoid local matching, a visual constraint algorithm is provided to eliminate mismatching point pairs so as to improve the matching precision of fingerprints, and the mismatching point pairs are removed according to the relation between the slopes and the lengths of the matching point pairs. The algorithm considers the complex conditions of loss, movement and the like of the minutiae, and combines with the mismatching points, so that the fingerprint identification performance is effectively improved.
The visual constraint fingerprint identification method based on the enhanced triangulation comprises the following steps:
step 1, preprocessing an input fingerprint picture to obtain a refined picture of the fingerprint, extracting features, obtaining a triangle set according to an EDT triangulation algorithm after extracting minutiae, and traversing domain minutiae determined by triangulation to form a triangle. Because two triangle sets are extracted, the two triangle sets have repeated triangles, and after the repeated triangles are removed, the two triangle sets are combined to form a final triangle set.
Step 2, after a final triangle set is obtained, calculating feature vectors of all triangles in the triangle set;
and 3, performing the operations of the step 1 and the step 2 on fingerprint images in the database, namely template fingerprint images. And judging whether triangles in the input fingerprint image are similar to the triangles in the template fingerprint image by using the feature vectors of the triangles, and when the similarity degree of the two triangles exceeds a set threshold, considering that the two triangles are successfully matched, and correspondingly matching the minutiae forming the two triangles so as to obtain a rough matching point set. Removing one-to-many or many-to-one matching point pairs in the coarse matching point set by using a double verification method, so that the matching points in the coarse matching point set are in one-to-one correspondence;
and 4, partial mismatching point pairs exist in the rough matching point set, and for each pair of corresponding point pairs in the rough matching point set, constructing adjacent local feature vectors to further verify the accuracy of the adjacent local feature vectors, so as to obtain a secondary matching point set. And because the feature points are locally matched, combining with the visual features, carrying out matching point rejection on the secondary matching point set again according to the slope and the length between the matching point pairs to obtain a final corresponding point set. And finally calculating a matching score, and judging whether the input fingerprint image and the template fingerprint image are successfully matched. And when the matching score is larger than the set threshold value, the matching of the two fingerprints is considered to be successful, otherwise, the matching is considered to be failed.
Further, the preprocessing of the input fingerprint picture comprises filtering, binarization and refinement.
The invention has the following beneficial effects:
the invention provides a visual constraint fingerprint identification method based on enhanced triangulation. The EDT-C method is improved, and a visual constraint method is provided. The EDT-C algorithm is improved by considering that the number of triangle sets obtained by the EDT-C is small and the information is also small. And processing the triangle set obtained by the EDT-C, taking one detail as an example, firstly taking out all triangles containing the minutiae, then obtaining a minutiae set composed of all minutiae (going to the minutiae) contained in the triangles, traversing the minutiae set, and randomly taking out two minutiae and the minutiae to form a triangle to form a new triangle set. And secondly, removing repeated triangles in the two triangle sets, and only leaving one triangle. And finally, merging the two triangle sets to obtain a final triangle set. Through the operation, more feature vectors can be obtained, so that the matching effect is better. In consideration of the mismatching point pair phenomenon in fingerprint matching, a visual constraint algorithm is provided, and the mismatching point pair is removed according to the relation between the slope and the length of the matching point sets to obtain a final matching point set. The visual constraint fingerprint identification method based on the enhanced triangulation considers the complex situations of losing, moving and the like of the minutiae, and effectively improves the fingerprint identification performance by combining with the method for eliminating mismatching points.
Drawings
Fig. 1 is a flow chart of fingerprint recognition according to the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The visual constraint fingerprint identification method based on the enhanced triangulation comprises the following steps:
step 1, preprocessing an input fingerprint picture to obtain a refined picture of the fingerprint, extracting features, obtaining a triangle set according to an EDT triangulation algorithm after extracting minutiae, and traversing the domain minutiae determined by triangulation to form a triangle by considering the existence of false minutiae and the fact that the triangulation algorithm is faster but has fewer obtained triangle features. Because two triangle sets are extracted, the two triangle sets have repeated triangles, and after the repeated triangles are removed, the two triangle sets are combined to form a final triangle set. The specific gravity of the triangle formed by the real minutiae points is increased through the operation, so that the accuracy of the characteristics is improved;
step 2, after a final triangle set is obtained, calculating feature vectors of all triangles in the triangle set;
and 3, performing the operations of the step 1 and the step 2 on fingerprint images in the database, namely template fingerprint images. And judging whether triangles in the input fingerprint image are similar to the triangles in the template fingerprint image by using the feature vectors of the triangles, and when the similarity degree of the two triangles exceeds a set threshold, considering that the two triangles are successfully matched, and correspondingly matching the minutiae forming the two triangles so as to obtain a rough matching point set. Removing one-to-many or many-to-one matching point pairs in the coarse matching point set by using a double verification method, so that the matching points in the coarse matching point set are in one-to-one correspondence;
and 4, partial mismatching point pairs exist in the rough matching point set, and for each pair of corresponding point pairs in the rough matching point set, constructing adjacent local feature vectors to further verify the accuracy of the adjacent local feature vectors, so as to obtain a secondary matching point set. And because the feature points are locally matched, combining with the visual features, carrying out matching point rejection on the secondary matching point set again according to the slope and the length between the matching point pairs to obtain a final corresponding point set. And finally calculating a matching score, and judging whether the input fingerprint image and the template fingerprint image are successfully matched. When the matching score is greater than a set threshold (the threshold is determined through experiments), the two fingerprints are considered to be successfully matched, otherwise, the matching is considered to be failed.
The EDT triangulation algorithm described in the step 1 specifically comprises the following steps:
EDT(P)＝DT(P)∪DT(P i )∪...∪DT(P m )
wherein: p represents a minutiae set consisting of m minutiae points extracted from an input fingerprint image, p= { (x) mi ,y mi ,θ mi )|i＝1,2,...,m},m i EDT (P) represents a triangle set obtained by EDT triangulation of the input fingerprint image minutiae set P; DT (P) means triangulating the set of minutiae points P to obtain a set of triangles; p (P) i Represents m i A set of minutiae points that participate in the composition of other minutiae points contained in the composition triangle; DT (P) i ) Representation pair P i Performing triangulation;
calculating feature vectors of the triangle in the step 2, wherein the feature vectors comprise geometric features and minutiae features; geometric features include side length and angle, assuming a triangle formed by (p i ,p j ,p k ) Three minutiae points consist of:
side length: calculation of p i p j ,p i p k ,p j p k The distances of three sides are ordered from big to small according to the length;
angle: three internal angles of the triangle are calculated, and the triangle is sequenced according to the sequence of the edge lengths;
minutiae characteristics: considering only the geometric features of triangles does not match triangles exactly, because triangles have translational rotation variations, and symmetrical matching is easy to occur. It is therefore necessary to add the directional field characteristics of minutiae points on the basis of geometric characteristics.
Assume that one triangle of the input fingerprint image is composed of (p i ,p j ,p k ) Three minutiae, then m= (θ) pi ,θ pj ,θ pk ) The method comprises the steps of carrying out a first treatment on the surface of the M represents the direction of three minutiae points in the triangle. Wherein θ is pi A direction field representing an i-th minutiae point of the input fingerprint image; wherein θ is pj A direction field representing a j-th minutiae point of the input fingerprint image; wherein θ is pk A direction field representing a kth minutiae point of the input fingerprint image;
because the minutiae orientation is the angle between the ridge line orientation and the coordinate axis, even for the same fingerprint, the orientation of the fingerprint minutiae will change after rotation. But the angular change between the minutiae directions is constant, using changeM p Representing the relative change between the minutiae directional fields, the calculation formula is as follows:
changeM p ＝(θ pj -θ pi ,θ pk -θ pj ,θ pi -θ pk )；；
wherein θ is pj -θ pi Represents p j And p i Is the difference between the directional fields of (a);
the double matching in the step 3 is specifically realized as follows:
it is assumed that a certain triangle in the input fingerprint image is defined by (p i ,p j ,p k ) Composition, respectively obtaining the side length characteristics (p ij ,p ik ,p jk ) Angle characteristic (alpha) pij ,α pik ,α pjk ) Minutiae feature changeM p The method comprises the steps of carrying out a first treatment on the surface of the Similarly, the same operation is performed on the template fingerprint image, assuming that a certain triangle in the template fingerprint image is formed by (q i ,q j ,q k ) Composition, respectively obtaining the side length characteristics (q ij ,q ik ,q jk ) Angle characteristic (alpha) qij ,α qik ,α qjk ) Minutiae feature changeM p The method comprises the steps of carrying out a first treatment on the surface of the The Euclidean distance of the corresponding characteristics of the two triangles is calculated respectively, and the specific judgment conditions are as follows:
3-1、
Diff dist ＝s(|p ij -q ij |)+s(|p ik -q ik |)+s(|p jk -q jk |),
3-2、
Diff ang ＝s(|α pij -α qij |)+s(|α pik -α qik |)+s(|α pjk -α qjk |),
3-3、Diff chang ＝s(|changeM p -changeM q |),
wherein the s function indicates that if the absolute value of the difference between a and b is smaller than the threshold value set in advance, s (|a-b|) is 1, and conversely, is 0; diff (Diff) dist Representing the number of differences between the corresponding side lengths, the absolute value of which is smaller than the threshold value; diff (Diff) ang Representing the number of the absolute values of the differences between the corresponding angles being smaller than a threshold value; diff (Diff) ang When the difference of the corresponding minutiae features is less than the threshold number; when Diff is dist ＝3，Diff ang ＝3，Diff chang =3, then consider that the two triangles match successfully; judging whether the detail points forming the triangle are matched or not through triangle matching; if triangle a is successfully matched by minutiae (p i ,p j ,p k ) Triangle B consists of (q i ,q j ,q k ) The composition is regarded as { p ] i ,q i },{p j ,q j },{p k ,q k Three minutiae matches, adding the corresponding points to the coarse matching point set RoughMathchingSet:
when adding matching point pairs into the coarse matching point set, firstly judging whether the coarse matching point set has corresponding matching point pairs, wherein the matchtime represents the matching scores of the corresponding minutiae points of the input fingerprint picture and the template fingerprint picture, if not, the matching points are directly added, the matchtime=1, and if so, the matching points are matched with the template fingerprint picture, and the original value is covered.
Because the minutiae-based point pattern matching algorithm only utilizes the local field features of minutiae, minutiae with similar features exist, after matching, minutiae in an input fingerprint image can be matched to a plurality of similar minutiae in a template fingerprint image, i.e. one-to-many conditions exist. Similarly, a plurality of minutiae points in the input fingerprint image will correspond to a plurality of similar minutiae points in the template fingerprint image. And matching is carried out by using a double matching method, so that minutiae in the input fingerprint image are successfully matched with only one minutiae in the template fingerprint image.
The matching is carried out by adopting a double matching method, and the specific steps are as follows:
let it be assumed that minutiae p in the input fingerprint image i The template minutiae set, namely, the minutiae set extracted from the template fingerprint image has a plurality of points corresponding to the minutiae set, namely:
matrichtime represents p i And q i Matching scores of (2); matchtime2 represents p i And q j And a matching score; matrichtime 3 represents p i And q k Matching scores of (2); i.e. p in the input fingerprint picture i Minutiae and q in template fingerprint pictures i ，q j ，q k The matching is successful, and the phenomenon of one-to-many exists; when this is the case, the minutiae corresponding to the maximum of matchtime, matchtime2, matchtime3 is taken as p i Is a matching point of (2); other matching point pairs are removed from the coarse matching point set. The same is done when there is a many-to-one case;
and 4, eliminating mismatching points by using a mode of combining field features and visual features, wherein the method is specifically realized as follows:
for the matching point pairs in the coarse matching point set, the correctness of the matching point pairs is verified by constructing adjacent local feature vectors,
(1) And for the matching point pair (f, g) in the rough matching point set, namely the f-th minutiae in the input fingerprint image and the g-th minutiae in the template fingerprint image are successfully matched. Where f and g are the f-th minutiae point of the input minutiae point set and the g-th minutiae point of the template minutiae point set, respectively. Firstly, the Euclidean distance between each minutiae in the input minutiae set and other minutiae in the input minutiae set is calculated, n minutiae closest to the minutiae f are taken, and adjacent local feature vectors of the minutiae f are constructed.
Loc(f)＝{(d 1 ,δ 1 ),(d 2 ,δ 2 ),...，(d l ,δ l ),...,(d n ,δ n )}，
Wherein d l Representing the Euclidean distance, delta, between a minutiae point f and its first closest minutiae point l Representing the difference in minutiae direction between the minutiae f and the surrounding first proximal minutiae.
For the minutiae g, firstly, calculating the Euclidean distance between each minutiae of the template minutiae set and other minutiae of the template minutiae set, taking n minutiae closest to the minutiae g, and constructing adjacent local feature vectors of the minutiae g; loc (g) = { (d) 1 ',δ 1 '),(d 2 ',δ 2 '),...,(d l ',δ l '),...,(d n ',δ n ')}
(2) Setting a flag bit whose initial value is 0, and verifying abs (d l -d l ')＜T1，abs(δ l -δ l ') < T2, wherein abs () represents taking absolute value; t1 and T2 are thresholds determined experimentally. When the absolute values of the difference between the distances and the difference between the angles of the matching point pairs are smaller than the set threshold values T1 and T2, the value of the flag is increased by one; otherwise, the value of the flag is unchanged. And after one matching point pair completes all verification, judging whether the value of the flag is larger than n/2, if the condition is met, considering f and g as the matched minutiae point pair passing the verification, and adding the matched minutiae point pair into a secondary matching point set.
(3) Repeating the steps (1) and (2) until all the matching point pairs in the coarse matching point set are verified, and obtaining a secondary matching point set.
(4) However, the partial matching is easy to succeed only by removing the mismatching points, the visual characteristics of the global minutiae are considered, and in general, the proportion of the mismatching point pairs in the total matching point pairs is smaller, but in extreme cases, the fingerprint image quality is poor or the nonlinear deformation is serious, so that the occupation of the mismatching point pairs in the total matching point pairs is larger; and calculating the distance and the slope between the matching point pairs, wherein for the secondary matching point set, the matching point pairs (w, v) are represented by w which are minutiae of the input fingerprint image, and the v is represented by minutiae of the template fingerprint image. The distance between the two minutiae points and the slope are calculated.
4-1、
(x w ,y w ) Represents the abscissa and the ordinate of the minutiae point w, (x) v ,y v ) Representing the abscissa and ordinate of minutiae point v, distance represents the distance between the calculated minutiae point w and minutiae point v;
4-2、
gradient represents the slope of the line connecting minutiae w and minutiae v;
4-3, respectively counting the number of positive values and negative values of the slope;
Poisitive number ＝sum(gradient(:)≥0)
Negative number ＝sum(gradient(:)＜0)
wherein gradient (:) represents the slope of all matching point pairs, and gradient (:) 0 represents the matching point pair with positive slope in all matching point pairs. Poisitive number Representing the number of the secondary matching point set matching minutiae pairs with the connecting line slope larger than or equal to 0; negative) number Representing the number of the secondary matching point set matching minutiae pairs with the connecting line slope smaller than 0;
solving for a reference gradient and length value:
Reference gradient representing the value of the reference gradient, when Poisitive number Greater than or equal to Negative number When the gradient is greater than or equal to 0, the median of the gradient is taken as the value of the reference gradient; otherwise, taking the ladderA median of the slopes having a degree less than 0 is taken as the value of the reference gradient; in the same way, the length value Reference is calculated distan 。
Diff gradient ＝abs(gradient-Re ference gradient )
Diff distan ＝abs(dis tan ce-Re ference distan )，
Diff gradient Representing absolute values of the current gradient and the reference gradient; diff (Diff) distan Representing absolute values of the current length and the reference length; when Diff is gradient ≤T g ，Diff distan ≤T d And adding the pair of matching points to the final matching point set, and eliminating the matching point set if not. T (T) d And T g Is a threshold determined experimentally.
Final calculation of matching scoreWhere matchnumber represents the number of minutiae that the input fingerprint image and the template fingerprint image eventually match, minute1 represents the total number of minutiae in the input fingerprint image, and minute2 represents the total number of minutiae in the template fingerprint image.
When the matching score is greater than a set threshold (the threshold is determined through experiments), the two fingerprints are considered to be successfully matched, otherwise, the matching is considered to be failed.
Claims (5)
1. The visual constraint fingerprint identification method based on the enhanced triangulation is characterized by comprising the following steps of:
firstly, preprocessing an input fingerprint picture to obtain a refined picture of the fingerprint, extracting features, obtaining a triangle set according to an EDT triangulation algorithm after extracting minutiae, and traversing domain minutiae determined by triangulation to form a triangle; because two triangle sets are extracted, the two triangle sets have repeated triangles, and after the repeated triangles are removed, the two triangle sets are combined to form a final triangle set;
step 2, after a final triangle set is obtained, calculating feature vectors of all triangles in the triangle set;
step 3, performing the operations of step 1 and step 2 on fingerprint images in the database, namely template fingerprint images; judging whether triangles in the input fingerprint image are similar to the triangles in the template fingerprint image or not by using the feature vectors of the triangles, and when the similarity degree of the two triangles exceeds a set threshold, considering that the two triangles are successfully matched, and correspondingly matching the minutiae forming the two triangles so as to obtain a rough matching point set; removing one-to-many or many-to-one matching point pairs in the coarse matching point set by using a double verification method, so that the matching points in the coarse matching point set are in one-to-one correspondence;
step 4, partial mismatching point pairs exist in the rough matching point set, and for each pair of corresponding point pairs in the rough matching point set, constructing adjacent local feature vectors to further verify the accuracy of the adjacent local feature vectors, so as to obtain a secondary matching point set; because the feature points are locally matched, combining with the visual features, carrying out matching point rejection on the secondary matching point set again according to the slope and the length between the matching point pairs to obtain a final corresponding point set; finally calculating a matching score, and judging whether the input fingerprint image and the template fingerprint image are successfully matched; when the matching score is larger than a set threshold value, the two fingerprints are considered to be successfully matched, otherwise, the matching is considered to be failed;
the EDT triangulation algorithm described in the step 1 specifically comprises the following steps:
EDT(P)＝DT(P)∪DT(P i )∪...∪DT(P m )
wherein: p represents a minutiae set consisting of m minutiae points extracted from an input fingerprint image, p= { (x) mi ,y mi ,θ mi )|i＝1,2,...,m},m i EDT (P) represents a triangle set obtained by EDT triangulation of the input fingerprint image minutiae set P; DT (P) means triangulating the set of minutiae points P to obtain a set of triangles; p (P) i Represents m i A set of minutiae points that participate in the composition of other minutiae points contained in the composition triangle; DT (P) i ) Representation pair P i Triangulation is performed.
2. The method for identifying the visual constraint fingerprint based on the enhanced triangulation according to claim 1, wherein the preprocessing of the input fingerprint picture comprises filtering, binarizing and refining.
3. The visual constraint fingerprint identification method based on enhanced triangulation according to claim 1 or 2, wherein step 2 calculates feature vectors of triangles, specifically as follows:
the feature vector contains geometric features and minutiae features; geometric features include side length and angle, assuming a triangle formed by (p i ,p j ,p k ) Three minutiae points consist of:
side length: calculation of p i p j ,p i p k ,p j p k The distances of three sides are ordered from big to small according to the length;
angle: three internal angles of the triangle are calculated, and the triangle is sequenced according to the sequence of the edge lengths;
minutiae characteristics:
assume that one triangle of the input fingerprint image is composed of (p i ,p j ,p k ) Three minutiae, then m= (θ) pi ,θ pj ,θ pk ) The method comprises the steps of carrying out a first treatment on the surface of the M represents the directions of three minutiae points in the triangle; wherein θ is pi A direction field representing an i-th minutiae point of the input fingerprint image; wherein θ is pj A direction field representing a j-th minutiae point of the input fingerprint image; wherein θ is pk A direction field representing a kth minutiae point of the input fingerprint image;
using changeM p Representing the relative change between the minutiae directional fields, the calculation formula is as follows:
changeM p ＝(θ pj -θ pi ,θ pk -θ pj ,θ pi -θ pk )；
wherein θ is pj -θ pi Represents p j And p i Is a difference between the directional fields of (a) and (b).
4. The visual constraint fingerprint identification method based on enhanced triangulation according to claim 3, wherein the double matching in the step 3 is specifically implemented as follows:
it is assumed that a certain triangle in the input fingerprint image is defined by (p i ,p j ,p k ) Composition, respectively obtaining the side length characteristics (p ij ,p ik ,p jk ) Angle characteristic (alpha) pij ,α pik ,α pjk ) Minutiae feature changeM p The method comprises the steps of carrying out a first treatment on the surface of the Similarly, the same operation is performed on the template fingerprint image, assuming that a certain triangle in the template fingerprint image is formed by (q i ,q j ,q k ) Composition, respectively obtaining the side length characteristics (q ij ,q ik ,q jk ) Angle characteristic (alpha) qij ,α qik ,α qjk ) Minutiae feature changeM p The method comprises the steps of carrying out a first treatment on the surface of the The Euclidean distance of the corresponding characteristics of the two triangles is calculated respectively, and the specific judgment conditions are as follows:
3-1、
3-2、
3-3、
wherein the s function indicates that if the absolute value of the difference between a and b is less than a threshold set in advance, s (|a-b|) is 1, and conversely, 0; diff (Diff) dist Representing the number of differences between the corresponding side lengths, the absolute value of which is smaller than the threshold value; diff (Diff) chang Representing the number of the absolute values of the differences between the corresponding angles being smaller than a threshold value; diff (Diff) ang Representing the number of differences between corresponding minutiae features less than a threshold; when Diff is dist ＝3，Diff ang ＝3，Diff chang =3, then consider that the two triangles match successfully; judging whether the detail points forming the triangle are matched or not through triangle matching; if triangle a is successfully matched by minutiae (p i ,p j ,p k ) Triangle B consists of (q i ,q j ,q k ) The composition is regarded as { p ] i ,q i },{p j ,q j },{p k ,q k Three minutiae matches, adding the corresponding points to the coarse matching point set RoughMathchingSet:
when adding matching point pairs into the coarse matching point set, firstly judging whether the coarse matching point set has corresponding matching point pairs, wherein the matchtime represents the matching scores of the corresponding minutiae points of the input fingerprint picture and the template fingerprint picture, if not, the matching points are directly added, the matchtime=1, and if so, the matching points are matched with the template fingerprint picture, and the original value is covered;
the matching is carried out by adopting a double matching method, and the specific steps are as follows:
let it be assumed that minutiae p in the input fingerprint image i The template minutiae set, namely, the minutiae set extracted from the template fingerprint image has a plurality of points corresponding to the minutiae set, namely:
matrichtime represents p i And q i Matching scores of (2); matchtime2 represents p i And q j Matching scores of (2); matrichtime 3 represents p i And q k Matching scores of (2); i.e. p in the input fingerprint picture i Minutiae and q in template fingerprint pictures i ，q j ，q k The matching is successful, and the phenomenon of one-to-many exists; when this is the case, the minutiae corresponding to the maximum of matchtime, matchtime2, matchtime3 is taken as p i Is a matching point of (2); removing other matching point pairs from the coarse matching point set; the same operation is also performed when there is a many-to-one case.
5. The enhanced triangulation-based visual constraint fingerprint identification method according to claim 4, wherein step 4, the mismatching points are removed by combining the neighborhood features and the visual features, and is specifically implemented as follows:
for the matching point pairs in the coarse matching point set, the correctness of the matching point pairs is verified by constructing adjacent local feature vectors,
(1) For the matching point pair (f, g) in the coarse matching point set, f and g are respectively the f-th minutiae of the input minutiae set and the g-th minutiae of the template minutiae set; firstly, calculating Euclidean distance between each minutiae in an input minutiae set and other minutiae in the input minutiae set, taking n minutiae closest to the minutiae f, and constructing adjacent local feature vectors of the minutiae f;
Loc(f)＝{(d 1 ,δ 1 ),(d 2 ,δ 2 ),...，(d l ,δ l ),...,(d n ,δ n )}，
wherein d l Representing the Euclidean distance, delta, between minutiae point f and its minutiae point closest to the first l Representing the difference in minutiae direction between the minutiae f and the surrounding first closest minutiae;
for the minutiae g, firstly, calculating the Euclidean distance between each minutiae of the template minutiae set and other minutiae of the template minutiae set, taking n minutiae closest to the minutiae g, and constructing adjacent local feature vectors of the minutiae g; loc (g) = { (d) 1 ',δ 1 '),(d 2 ',δ 2 '),...,(d l ',δ l '),...,(d n ',δ n ')}
(2) Setting a flag bit whose initial value is 0, and verifying abs (d l -d l ')＜T1，abs(δ l -δ l ') < T2, wherein abs () represents taking absolute value; t1 and T2 are experimentally determined thresholds; when the absolute values of the difference between the distances and the difference between the angles of the matching point pairs are smaller than the set threshold values T1 and T2, the value of the flag is increased by one; otherwise, the value of the flag is unchanged; after one matching point pair completes all verification, judging whether the value of the flag is larger than n/2, if the value of the flag is larger than n/2, considering f and g as the matched minutiae point pair passing the verification, and adding the matching minutiae point pair into a secondary matching point set;
(3) Repeating the steps (1) and (2) until all the matching point pairs in the coarse matching point set are verified, and obtaining a secondary matching point set;
(4) Calculating the distance and the slope between the matching point pairs, wherein for the secondary matching point set, the matching point pairs (w, v), w represents the minutiae of the input fingerprint image, and v represents the minutiae of the template fingerprint image; calculating the distance and the slope between the two detail points;
4-1、
(x w ,y w ) Represents the abscissa and the ordinate of the minutiae point w, (x) v ,y v ) Representing the abscissa and ordinate of minutiae point v, distance represents the distance between the calculated minutiae point w and minutiae point v;
4-2、
gradient represents the slope of the line connecting minutiae w and minutiae v;
4-3, respectively counting the number of positive values and negative values of the slope;
Poisitive number ＝sum(gradient(:)≥0)
Negative number ＝sum(gradient(:)＜0)
wherein gradient (:) represents the slope of the matching point pair, and gradient (:) is equal to or greater than 0 represents the matching point pair with positive slope in the matching point pair; poisitive number Representing the number of the secondary matching point set matching minutiae pairs with the connecting line slope larger than or equal to 0; negative) number Representing the number of the secondary matching point set matching minutiae pairs with the connecting line slope smaller than 0;
solving for a reference gradient and length value:
Reference gradient representing the value of the reference gradient, when Poisitive number Greater than or equal to Negative number When the gradient is greater than or equal to 0, the median of the gradient is taken as the value of the reference gradient; otherwise, taking the median of the gradient smaller than 0 slope as the value of the reference gradient; in the same way, the length value Reference is calculated distan ；
Diff gradient ＝abs(gradient-Reference gradient )
Diff distan ＝abs(distance-Reference distan )，
Diff gradient Representing absolute values of the current gradient and the reference gradient; diff (Diff) distan Representing absolute values of the current length and the reference length; when Diff is gradient ≤T g ,Diff distan ≤T d When the matching points are matched, the matching point pairs are added into the final matching point set, and otherwise, the matching point pairs are removed; t (T) d And T g Is a threshold determined experimentally;
final calculation of matching scoreWherein matchnumber represents the number of minutiae that the input fingerprint image and the template fingerprint image are finally matched, and minutiae 1 represents the total number of minutiae in the input fingerprint image and minutiae 2 represents the total number of minutiae in the template fingerprint image;
and when the matching score is larger than the set threshold value, the matching of the two fingerprints is considered to be successful, otherwise, the matching is considered to be failed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011209604.9A CN112487867B (en) | 2020-11-03 | 2020-11-03 | Visual constraint fingerprint identification method based on enhanced triangulation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011209604.9A CN112487867B (en) | 2020-11-03 | 2020-11-03 | Visual constraint fingerprint identification method based on enhanced triangulation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112487867A CN112487867A (en) | 2021-03-12 |
| CN112487867B true CN112487867B (en) | 2024-04-12 |
Family
ID=74927710
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011209604.9A Active CN112487867B (en) | 2020-11-03 | 2020-11-03 | Visual constraint fingerprint identification method based on enhanced triangulation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112487867B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114821685B (en) * | 2022-05-12 | 2023-05-02 | 黑龙江省科学院智能制造研究所 | PSO (particle swarm optimization) optimization Delaunay triangulation-based non-contact 3D fingerprint identification method |
| CN115497125B (en) * | 2022-11-17 | 2023-03-10 | 上海海栎创科技股份有限公司 | Fingerprint identification method, system, computer equipment and computer readable storage medium |
| CN116386091B (en) * | 2022-11-18 | 2024-04-02 | 荣耀终端有限公司 | Fingerprint identification method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1912889A (en) * | 2005-08-11 | 2007-02-14 | 中国科学院自动化研究所 | Deformed fingerprint identification method based on local triangle structure characteristic collection |
| CN102262730A (en) * | 2011-08-15 | 2011-11-30 | 山东志华信息科技股份有限公司 | Fingerprint matching method based on multiple reference point pairs |
| CN105787451A (en) * | 2016-02-29 | 2016-07-20 | 南京邮电大学 | Fingerprint matching method based on multi-judgment point mode |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104156707B (en) * | 2014-08-14 | 2017-09-22 | 深圳市汇顶科技股份有限公司 | Fingerprint identification method and its fingerprint identification device |
-
2020
- 2020-11-03 CN CN202011209604.9A patent/CN112487867B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1912889A (en) * | 2005-08-11 | 2007-02-14 | 中国科学院自动化研究所 | Deformed fingerprint identification method based on local triangle structure characteristic collection |
| CN102262730A (en) * | 2011-08-15 | 2011-11-30 | 山东志华信息科技股份有限公司 | Fingerprint matching method based on multiple reference point pairs |
| CN105787451A (en) * | 2016-02-29 | 2016-07-20 | 南京邮电大学 | Fingerprint matching method based on multi-judgment point mode |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112487867A (en) | 2021-03-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112487867B (en) | Visual constraint fingerprint identification method based on enhanced triangulation | |
| Valdes-Ramirez et al. | A Review of Fingerprint Feature Representations and Their Applications for Latent Fingerprint Identification: Trends and Evaluation. | |
| CN107748877B (en) | Fingerprint image identification method based on minutiae and textural features | |
| CN110097093B (en) | Method for accurately matching heterogeneous images | |
| Prasad et al. | A survey of fingerprint recognition systems and their applications | |
| US7359532B2 (en) | Fingerprint minutiae matching using scoring techniques | |
| CN107958443B (en) | Fingerprint image splicing method based on ridge line characteristics and TPS deformation model | |
| US8781177B2 (en) | Method and computer program product for extracting feature vectors from a palm image | |
| CN101114335A (en) | Full angle rapid fingerprint identification method | |
| Naderi et al. | Fusing iris, palmprint and fingerprint in a multi-biometric recognition system | |
| Fei et al. | Enhanced minutiae extraction for high-resolution palmprint recognition | |
| CN111914755A (en) | Eight-direction gradient-solving fingerprint identification model | |
| Akbar et al. | Palm vein biometric identification system using local derivative pattern | |
| Więcław | A minutiae-based matching algorithms in fingerprint recognition systems | |
| WO2016011640A1 (en) | Identification method based on handprint imaging | |
| CN108664940A (en) | A kind of partial fingerprints matching process and system | |
| Fei et al. | Fingerprint singular points extraction based on the properties of orientation model | |
| You et al. | Finger vein recognition method based on gaussian low-pass filter and direction detection | |
| Mistry et al. | Computationally efficient vanishing point detection algorithm based road segmentation in road images | |
| Li et al. | Multi-feature score fusion for fingerprint recognition based on neighbor minutiae boost | |
| Elmouhtadi et al. | Fingerprint indexing based barycenter triangulation | |
| CN116665256B (en) | Fingerprint matching method based on fingerprint image local area quality | |
| CN115050017B (en) | Electric power fitting surface inscription identification method based on machine vision | |
| CN113096168B (en) | Optical remote sensing image registration method and system combining SIFT points and control line pairs | |
| Mazaheri et al. | Fingerprint matching using an onion layer algorithm of computational geometry based on level 3 features |
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 |