CN108268836A - Automatic fingerprint classification method - Google Patents

Automatic fingerprint classification method Download PDF

Info

Publication number
CN108268836A
CN108268836A CN201711490895.1A CN201711490895A CN108268836A CN 108268836 A CN108268836 A CN 108268836A CN 201711490895 A CN201711490895 A CN 201711490895A CN 108268836 A CN108268836 A CN 108268836A
Authority
CN
China
Prior art keywords
block
type
row
fingerprint
judged
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
Application number
CN201711490895.1A
Other languages
Chinese (zh)
Other versions
CN108268836B (en
Inventor
刘汉英
邓昀
周剑勋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guilin University of Technology
Original Assignee
Guilin University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guilin University of Technology filed Critical Guilin University of Technology
Priority to CN201711490895.1A priority Critical patent/CN108268836B/en
Publication of CN108268836A publication Critical patent/CN108268836A/en
Application granted granted Critical
Publication of CN108268836B publication Critical patent/CN108268836B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/12—Fingerprints or palmprints
    • G06V40/1365—Matching; Classification
    • 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
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)
  • Collating Specific Patterns (AREA)

Abstract

The invention discloses a kind of automatic fingerprint classification methods.Fingerprint is divided into six classes by the present invention：Left dustpan type, right dustpan type, bucket type, arch form, cusped arch type and it can not judge.This fingerprint classification depends on fingerprint-block pattern features and fingerprint singularity（Central point and triangulation point）Number and location.Analyzing and training fingerprint image Block direction field generates taxonomy database（4 direction block numbers, 4 direction position of direction block, 8 direction position of direction block can not judge database）.To test fingerprint image direction field, classified according to taxonomy database.The method of the present invention is not easy to be influenced by fingerprint quality, to the fingerprint image that can not be judged, according to singular point quantity, position and taxonomy database, judges again.

Description

Automatic fingerprint classification method
Technical field
The present invention relates to field of computer technology, more particularly to a kind of automatic fingerprint classification method.
Background technology
Everyone fingerprint is unique, is frequently utilized for the identity of identification people.Fingerprint recognition need fingerprint base The fingerprint specified is found out in a large amount of fingerprints.In order to reduce seeking scope, first classify to fingerprint, then to same category of fingerprint It is compared, accelerates recognition speed.
Following steps are passed through in general fingerprint classification：Fingerprint segmentation, the calculation block field of direction find singular point and fingerprint point Class.
First, fingerprint segmentation
Original fingerprint image (Fig. 1) is divided into foreground and background two parts, only prospect is handled.Segmentation will usually refer to Print image is divided into the identical fritter of size, is divided according to features such as the gray scale of image, variance, orientation consistency, contrasts, Generation segmentation mask.
2nd, the field of direction is calculated
According to the grey scale change of fingerprint image, section technique direction, generation block directed graph (Fig. 2).
3rd, singular point is found
The positioning generally use direction revolving property (Poincare indexes) of singular point：In certain point field, around the point Closed curve rotate a circle, calculate the variable quantity summation that rotates of direction, the direction change amount of central point is 180, triangulation point Direction variable is -180, other positions 0.
4th, fingerprint classification
Henry classification modes are according to singular point position and quantity, the field of direction, Ridge following and textural characteristics, by fingerprint point Into five classes：Left dustpan type (left loop), right dustpan type (right loop), bucket type (whorl), arch form (arch) and cusped arch type (tented arch).They have following characteristics：
Left loop line enters from left side and goes out from left side, has 1 central point and 1 triangulation point (it is right to be located at central point Side)；
Right loop line enters from right side and goes out from right side, has 1 central point and 1 triangulation point (it is left to be located at central point Side)；
A bucket type at least streak line has 360 degree rotation, there is 2 central points and 2 triangulation points；
A part of streakline of arch form enters from fingerprint side, goes out from opposite side, without singular point；
Cusped arch type is similar with arch form, and at least a streak line has obvious cusped arch, there is 1 central point and 1 triangle Point (is located at immediately below singular point).
Fingerprint quality is relatively low or fingerprint is not full-time, and the inflection point detection of fingerprint can be relatively difficult or can not be detected, and does not look for It cannot directly be determined as arch form to singular point, only rely on unusual point feature and be easy to cause erroneous judgement.At present, this problem is not fine It solves.
Invention content
The object of the present invention is to provide a kind of automatic fingerprint classification methods.This method is not easy to be influenced by fingerprint quality, Fingerprint is divided into five classes according to block directed graph feature and fingerprint singularity (central point and triangulation point) number and location and can not be sentenced Determine class.
In order to reach this purpose, the present invention provides a kind of automatic fingerprint classification methods, include the following steps, wherein： (1)~(4) step be training process (as shown in Figure 4), for generate taxonomy database (4 direction block numbers, 4 direction position of direction block, 8 direction position of direction block, can not judge database), taxonomy database may be updated, and (5)~(13) step is test process, such as Shown in Fig. 5：
(1) training is manually divided into six classes with fingerprint image, the fingerprint for being judged to first five class is trained.By the finger of H × W Print image is divided into the block of 16 × 16 sizes, and image shares bi × bj blocks, calculation block pattern features (4 direction block numbers, direction block 8 directions position), by (14 row, the 1st~4 row difference table in block directed graph feature and fingerprint pattern deposit feature array feature Show：1 block number of direction, 2 block number of direction, 3 block number of direction, 4 block number of direction, 8 direction positions of the 5th~13 row each piece of barycenter of storage, 14th is classified as fingerprint pattern)；
(2) block directed graph feature in feature array feature (4 direction block number) and fingerprint pattern are taken, deletes and repeats to go, delete Except feature (4 direction block number) identical but different types of row, (5 row, divide 4 direction block number database class_dire_num of generation Not Wei 1 block number of direction, 2 block number of direction, 3 block number of direction, 4 block number of direction, fingerprint pattern)；
(3) feature array feature is taken, deletes the row that can classify by 4 direction block numbers, according to 8 direction position of direction block, 4 direction position (quadrant) of outgoing direction block is calculated, deletes and repeats to go, delete feature (4 direction block numbers, 4 direction position of direction block) phase Same but different types of row, (14 row, the 1st~4 row represent 4 direction location database class_dire4 of generation direction block respectively： 1 block number of direction, 2 block number of direction, 3 block number of direction, 4 block number of direction, 4 direction positions of the 5th~13 row each piece of barycenter of storage, the 14th It is classified as fingerprint pattern)；
(4) feature array feature is taken, deletes the row that can classify by 4 direction block numbers, 4 direction position of direction block, is deleted It repeats to go, deletes feature (4 direction block numbers, 8 direction position of direction block) identical but different types of row, 8 direction of generation direction block (14 row, the 1st~4 row represent location database class_dire8 respectively：1 block number of direction, 2 block number of direction, 3 block number of direction, side To 4 block numbers, 8 direction positions of the 5th~13 row each piece of barycenter of storage, the 14th is classified as fingerprint pattern), feature (4 direction block numbers, side To 8 direction position of block) generation of identical but different types of row can not judge database cantjudge (the same class_ of form dire8)；
(5) test fingerprint is pre-processed, calculation block pattern features (4 direction block numbers, 8 direction position of direction block)；
(6) 4 direction block number database class_dire_num are searched for 4 direction block numbers of test fingerprint, if it is found, defeated Go out fingerprint pattern, test terminates, and otherwise performs step (7)；
(7) according to 8 direction position of the direction block of test fingerprint, calculate 4 direction position of outgoing direction block, with 4 direction block numbers and 4 direction position direction of search block of direction block, 4 direction location database class_dire4, if it is found, output fingerprint pattern, is surveyed Examination terminates, and otherwise performs step (8)；
(8) with 8 direction location database class_dire8 of 4 direction block numbers and 8 direction position direction of search block of direction block, If it is found, output fingerprint pattern, test terminate, step (9) is otherwise performed；
(9) database cantjudge can not be judged with 4 direction block numbers and 8 direction location finding of direction block, if if found Dry row, it would be possible in type deposit type array lx, otherwise remember lx (1)=0, represent in training fingerprint without there is this feature Fingerprint；
(10) singular point, block position and number where record singular point are found；
(11) pseudo- singular point is deleted；
(12) if not finding singular point, expanded scope searches for a central point；
(13) judge fingerprint pattern by Henry sorting techniques (singular point number, position) and type array lx, can not judge Be judged to can not judge.
Six classes are 1. left dustpan types in described (1) step, and 2. right dustpan types, 3. bucket types, 4. arch forms, 5. cusped arch types and 6. can not sentence It is disconnected；
The calculating of block directed graph feature includes the following steps：
A. block directed graph dirbo, section technique mask fmsk are calculated using gradient method；
B. four direction feature (such as Fig. 6 shows) is taken, block directed graph dirbo is changed to four directional diagram dir4 of block (such as Fig. 7 shows), Method is
C. directional diagram 1 is generated by four directional diagram dir4, with collar plate shape structural element corrosion position Fig. 1 that radius is 1, calculated Corrode rear direction Fig. 1 (such as Fig. 8 (a)) connection component, calculate connection component barycenter 8 direction position pos8 (1-8 represent, (number of pixel is less than or equal to ignore this block when 4 in connection component) as shown in Figure 9)；Similarly calculate direction 2,3,4 block number and Centroid position (such as Fig. 8 (b), 8 (c), 8 (d) show), is stored in feature array feature, each fingerprint a line.
In described (3) step according to 8 direction position pos8 of direction block calculate direction block 4 direction position pos4 (1~4 represents, Respectively represent 1~4 quadrant, as shown in Figure 10)) method be
Singular point includes central point and triangulation point in described (10) step, includes the following steps：
A. singular point central point number singularcore and triangulation point number singulardelta are initialized as 0；
B. to removing image border (the first row, last column, first row, last row) in fingerprint block directed graph dirbo Each outer square calculates the variable quantity summation Poincare that direction rotates (such as Fig. 9 shows)：
△dk=dk- d(k+1)mod8K=0,1 ... 7
If △ dk>=90, △ dk=△ dk- 180
If △ dk≤ -90, △ dk=△ dk+180
C. such as fruit block (i, j), block (i-1, j), block (i-1, j-1), the Poincare values of block (i, j-1) are 180, Block (i, j) is foreground blocks, and it is foreground blocks that its 3 × 3 neighborhood, which has 5 pieces or more, then central point number singularcore adds 1, will Position i, j are recorded in position array xx1 and yy1；
D. such as fruit block (i, j), block (i-1, j) or block (i-1, j-2), block (i-1, j-1), block (i, j-1) Poincare values are -180, and block (i, j) is foreground blocks, and it is foreground blocks that its 3 × 3 neighborhood, which has 5 pieces or more, triangulation point number Singulardelta is 0 or singulardelta is 1 and distance is more than 3 between previous triangulation point, then triangulation point number Singulardelta adds 1, position is recorded i, j is in position array xx2 and yy2.
It is as follows that described (11) step deletes pseudo- unusual point methods：
A. to each block containing central point, if in its 5 × 5 field containing Poincare values be -180 block, Subtract 1 without central point, central point number singularcore in the block, the corresponding positions recorded in delete position array xx1 and yy1 It puts；
B. to each block containing triangulation point, if containing the block that Poincare values are 180 in its 5 × 5 field, it should Subtract 1 without triangulation point, triangulation point number singulardelta in block, the corresponding positions recorded in delete position array xx2 and yy2 It puts；
Described (12) step searches for a central point, and method is as follows：
To Poincare values in (10) step result for 180 and not image border (the 1st, 2 row, 1,2 rows reciprocal, the 1st, 2 Row, inverse 1,2 arrange) block (i, j), calculate the variable quantity summation Poincare that rotates of direction by Figure 11 (b)：
△dk=dk- d(k+1)mod8K=0,1 ... 7
If △ dk>=90, △ dk=△ dk- 180
If △ dk≤ -90, △ dk=△ dk+180
If Poincare (i, j)=180, position i, j are recorded in position by central point number singularcore=1 In array xx1 and yy1, step (13) is performed.
Described (13) step sorting technique is as follows：
A. if central point number singularcore=2 or triangulation point number singulardelta=2, and number of types Containing 3 or lx (1)=0 in group lx, it is judged to bucket type；
B. if central point number singularcore=1 or triangulation point number singulardelta=1, type array There are two element, and one of them is 4, then is judged to another type；
C. if central point number singularcore=0 and triangulation point number singulardelta=0, and number of types Containing element 4 in group, it is judged to arch form；
D. if central point number singularcore=1 and triangulation point number singulardelta=0, by with lower section Method judges：
It is not arch form, but can not judge if central point bottom left section or lower right-most portion prospect are less than 5 pieces；
If putting bottom left section centered on s1 other than left margin and lower boundary, the block number more than -90 degree and less than 0 degree, s2 Centered on point lower right-most portion other than right margin and lower boundary, more than 0 degree and be less than 90 degree of block number；
If s1<2, and contain 1 in type array, it is judged to left dustpan type；
If s2<2, and contain 2 in type array, it is judged to right dustpan type；
If s1-s2>Thre1, s2<Thre2, and contain 2 in type array, it is judged to right dustpan type；
If s2-s1>Thre3, s1<Thre4, and contain 1 in type array, it is judged to left dustpan type；
If s1>Thre5, s2>Thre5, and contain 3 in type array, it is judged to bucket type；
Thre1, thre2, thre3, thre4, thre5 are threshold value；
E. if central point number singularcore=0 and triangulation point number singulardelta=1, triangle is calculated Relative position yy2 (1)/bj of the point position in figure (bj is picture traverse, units chunk)；
If yy2/bj>Thre6, and contain 1 in type array, it is judged to left dustpan type；
If yy2/bj<Thre7, and contain 2 in type array, it is judged to right dustpan type；
Thre6, thre7 are threshold value；
F. if central point number singularcore=1 and triangulation point number singulardelta=1, center is calculated Point and triangulation point line and center line angleJudge by the following method：
If π/12>slope>- π/12, and contain 5 in type array, it is judged to cusped arch type；
If π/12>slope>- π/12, and 5 are not contained in type array, if putting bottom left section centered on s1 in addition to a left side Outside boundary and lower boundary, 0 degree of block number is spent and is less than more than -90, and point lower right-most portion is in addition to right margin and lower boundary centered on s2 Outside, the block number more than 0 degree and less than 90 degree；
If s1<2, and contain 1 in type array, it is judged to left dustpan type；
If s2<2, and contain 2 in type array, it is judged to right dustpan type；
If s1-s2>Thre1, s2<Thre2, and contain 2 in type array, it is judged to right dustpan type；
If s2-s1>Thre3, s1<Thre4, and contain 1 in type array, it is judged to left dustpan type；
If s1>Thre5, s2>Thre5, and contain 3 in type array, it is judged to bucket type；
Thre1, thre2, thre3, thre4, thre5 are threshold value with step d；
If slope >=π/12, and contain 1 in type array, it is judged to left dustpan type；
If slope≤- π/12, and contain 2 in type array, it is judged to right dustpan type；
G. it if being unsatisfactory for a, b, c, d, e, f, is judged to not judge.
Description of the drawings
Fig. 1 is original fingerprint image.
Fig. 2 is block directed graph.
Fig. 3 is mask, and prospect is white, and background is black.
Fig. 4 is the flow chart of training process.
Fig. 5 is the flow chart of test process.
Fig. 6 is the four direction of directional diagram.
Fig. 7 is four directional diagrams, and direction 4 is arrived in direction 1, and direction 1 is black, and color is gradually thin out, and direction 4 is light gray, is carried on the back Scape is white.
Fig. 8 is all directions figure, wherein (a) is directional diagram 1, (b) directional diagram 2, (c) directional diagram 3, (d) directional diagram 4.
Fig. 9 is 8 direction positions.
Figure 10 is 4 direction positions.
Figure 11 is direction of fingerprint revolving property, wherein (a) is 3 × 3 neighborhoods, (b) is 5 × 5 neighborhoods.
Specific embodiment
Embodiment：
The flow of the present invention as shown in Fig. 2, training is numbered with fingerprint image, to be manually determined as the fingerprint of first five class into Row training, training process are as follows：
The fingerprint image of 1.H × W is divided into the block of 16 × 16 sizes, block diagram image heightBlock diagram image width isTo each 16 × 16 piecemeal, calculate Block direction and mask, method are as follows：
The average gray of block
Block standard deviation
Block grey-scale contrast(n1 is that gray value is greater than or equal to block gray scale in block The points of mean value avgb, n2 are the points that gray value is less than block gray average avgb in block, and t1 is that gray value is more than or waits in block In the sum of all the points gray value of block gray average avgb, t2 is all the points ash of the gray value less than block gray average avgb in block The sum of angle value)；
Sx,SyFor Sobel operators；
Block direction consistency
Block directionBlock directed graph such as Fig. 2 shows；
Work as avgb<Thre6, varb<Thre7, zb<Thre8, cohb<It is background block during thre9, is otherwise foreground blocks, Thre6, thre7, thre8, thre9 are threshold value, and mask fmsk such as Fig. 3 shows.
2. calculation block pattern features, method are as follows：
Four direction of blockFour directional diagram of block is as shown in Figure 7；
Directional diagram 1：
Structural element：
Diag1=diag1 Θ se；Diag1 is corroded using se, shown in corrosion rear direction Fig. 1 such as Fig. 8 (a)；
If the number n1 of pixel is more than 5 in connected component, 8 direction position pos8 of connected component barycenter, side are calculated Method is：
Xi, yi are the coordinate of pixel (i.e. block)；
Similarly calculate directional diagram 2, directional diagram 3, directional diagram 4；
Directional diagram 2：Diag2 is corroded using se, corrodes rear To Fig. 2 such as Fig. 8 (b) Suo Shi, 2 block number n2=2 of directional diagram；
Directional diagram 3：Diag3 is corroded using se, corrosion rear direction Fig. 3 As shown in Fig. 8 (c), 3 block number n3=2 of direction (the number n of pixel is less than or equal to not include when 4 in connected component)；
Directional diagram 4：Diag4 is corroded using se, corrodes rear To Fig. 4 such as Fig. 8 (d) Suo Shi, 2 block number n4=2 of directional diagram.
3. generate feature array feature
Feature (k, 1)=n1；Feature (k, 2)=n2；Feature (k, 3)=n3；Feature (k, 4)=n4；
Feature (k, 14)=fingerprint pattern (1~5)；8 direction positions of the 5th~13 row each piece of barycenter of storage；
K is that training fingerprint is numbered；
4. generating 4 direction block number database class_dire_num, method is as follows：
Feature array feature the 1st~4 is taken to arrange, the 14th row, generation array class_dire_num (totally 5 row)；
Array class_dire_num is deleted to repeat to go；
It is identical to delete the 1st~4 row of array class_dire_num, but the row that the 5th row are different.
5. generating 4 direction location database class_dire4 of direction block, method is as follows：
Copy feature array：Class_dire4=feature；
The row that can classify by 4 direction block numbers is deleted, method is to delete the 1st~4 row and 4 direction blocks in class_dire4 The identical row of 1st~4 row in each rows of number database class_dire_num；
4 direction position of direction block is calculated, method is that the 5th~13 columns value pos8 in class_dire4 is converted：
Array class_dire4 is deleted to repeat to go；
Deletion array class_dire4 the 1st~13 is identical, but the row that the 14th row are different.
6. it generates 8 direction location database class_dire8 of direction block and can not judge database cantjudge, method It is as follows：
Copy feature array：Class_dire8=feature；
The row that can classify by 4 direction block numbers is deleted, method is to delete the 1st~4 row and 4 direction blocks in class_dire8 The identical row of 1st~4 row in each rows of number database class_dire_num；
The row that can classify by 4 direction positions is deleted, method is：
The 5th~13 columns value pos8 is converted in class_dire8：It is stored in interim array In class_dire8_temp；
Record the 1st~13 row and 4 direction location database class_ of direction block in interim array class_dire8_temp The line number id of the identical row of 1st~13 row in each rows of dire4；
Delete the row that line number in class_dire8 is id；
Array class_dire8 is deleted to repeat to go；
Replicate array：Cantjudge=class_dire8；
Deletion array class_dire8 the 1st~13 is identical, but the row that the 14th row are different, obtains 8 direction positional number of direction block According to library class_dire8.
Array cantjudge removes the row that class_dire8 contains, and obtains not judging database cantjudge.
Test process is as follows：
1. a pair test fingerprint pre-processes, calculation block pattern features (4 direction block numbers, 8 direction position of direction block), side Method is the same as training process step 1, step 2；
2. according to 4 direction block number n1, n2, n3, n4 search for 4 direction block number database class_dire_num the 1st~4 row, If it is found, exporting the row the 5th row, i.e. fingerprint pattern, test terminates, and otherwise performs step 3；
3. block 8 direction position pos8 in direction is converted：According to 4 direction block number n1, n2, n3, N4 and direction block 4 direction position pos4,4 direction location database class_dire4 the 1st~13 of the direction of search block row, if looked for It arrives, exports the row the 14th row, is i.e. fingerprint pattern, test terminates, and otherwise performs step 4；
4. according to 4 direction block number n1, n2, n3,8 direction position pos8 of n4 and direction block, 8 direction positional number of direction of search block It is arranged according to library class_dire8 the 1st~13, if it is found, exporting the row the 14th row, i.e. fingerprint pattern, test terminates, otherwise perform Step 5；
5. according to 4 direction block number n1, n2, n3,8 direction position pos8 of n4 and direction block, search can not judge database If cantjudge finds several rows, it will be expert in the 14th row deposit type array lx, otherwise remember lx (1)=0；
6. finding singular point, block position and number, method are where record singular point：
A. singular point central point number singularcore and triangulation point number singulardelta are initialized as 0；
B. to removing image border (the first row, last column, first row, last row) in fingerprint block directed graph dirbo Each outer square calculates the variable quantity summation Poincare that direction rotates (such as Figure 11 (a) shows)：
△dk=dk- d(k+1)mod8K=0,1 ... 7
If △ dk>=90, △ dk=△ dk- 180
If △ dk≤ -90, △ dk=△ dk+180
C. such as fruit block (i, j), block (i-1, j), block (i-1, j-1), the Poincare values of block (i, j-1) are 180, Block (i, j) is foreground blocks, and it is foreground blocks that its 3 × 3 neighborhood, which has 5 pieces or more, and central point number singularcore adds 1, by position I is put, j is recorded in position array xx1 and yy1；
D. such as fruit block (i, j), block (i-1, j) or block (i-1, j-2), block (i-1, j-1), block (i, j-1) Poincare values are -180, and block (i, j) is foreground blocks, and it is foreground blocks that its 3 × 3 neighborhood, which has 5 pieces or more, triangulation point number Singulardelta is 0 or singulardelta is 1 and distance is more than 3 between previous triangulation point, triangulation point number Singulardelta adds 1, position is recorded i, j is in position array xx2 and yy2.
7. deleting pseudo- singular point, method is：
A. to each block containing central point, if in its 5 × 5 field containing Poincare values be -180 block, Subtract 1 without central point, central point number singularcore in the block, the corresponding positions recorded in delete position array xx1 and yy1 It puts；
B. to each block containing triangulation point, if containing the block that Poincare values are 180 in its 5 × 5 field, it should Subtract 1 without triangulation point, triangulation point number singulardelta in block, the corresponding positions recorded in delete position array xx2 and yy2 It puts；
8. if do not find singular point, expanded scope searches for a central point, and method is：
To Poincare values in the 6th step result for 180 and not image border (the 1st, 2 row, 1,2 rows reciprocal, the 1st, 2 row, Inverse 1,2 arranges) block (i, j), the variable quantity summation Poincare that rotates of direction calculated by Figure 11 (b)：
△dk=dk- d(k+1)mod8K=0,1 ... 7
If △ dk>=90, △ dk=△ dk- 180
If △ dk≤ -90, △ dk=△ dk+180
If Poincare (i, j)=180, central point number singularcore=1, block position where writing down central point I, j are recorded in position array xx1 and yy1, perform step 9.
9. judging fingerprint pattern by Henry sorting techniques (singular point number, position) and type array lx, can not judge It is judged to not judge, method is：
A. if central point number singularcore=2 or triangulation point number singulardelta=2, and number of types Containing 3 or lx (1)=0 in group lx, it is judged to bucket type；
B. if central point number singularcore=1 or triangulation point number singulardelta=1, type array There are two element, and one of them is 4, then is judged to another type；
C. if central point number singularcore=0 and triangulation point number singulardelta=0, and number of types Containing element 4 in group, it is judged to arch form；
D. if central point number singularcore=1 and triangulation point number singulardelta=0, by with lower section Method judges：
It is not arch form, but can not judge if central point bottom left section or lower right-most portion prospect are less than 5 pieces；
If putting bottom left section centered on s1 other than left margin and lower boundary, the block number more than -90 degree and less than 0 degree, s2 Centered on point lower right-most portion other than right margin and lower boundary, more than 0 degree and be less than 90 degree of block number；
If s1<2, and contain 1 in type array, it is judged to left dustpan type；
If s2<2, and contain 2 in type array, it is judged to right dustpan type；
If s1-s2>Thre1, s2<Thre2, and contain 2 in type array, it is judged to right dustpan type；
If s2-s1>Thre3, s1<Thre4, and contain 1 in type array, it is judged to left dustpan type；
If s1>Thre5, s2>Thre5, and contain 3 in type array, it is judged to bucket type；
Thre1, thre2, thre3, thre4, thre5 are threshold value；
E. if central point number singularcore=0 and triangulation point number singulardelta=1, triangle is calculated Relative position yy2 (1)/bj of the point position in figure (bj is picture traverse, units chunk)；
If yy2/bj>Thre6, and contain 1 in type array, it is judged to left dustpan type；
If yy2/bj<Thre7, and contain 2 in type array, it is judged to right dustpan type；
Thre6, thre7 are threshold value；
F. if central point number singularcore=1 and triangulation point number singulardelta=1, center is calculated Point and triangulation point line and center line angleJudge by the following method：
If π/12>slope>- π/12, and contain 5 in type array, it is judged to cusped arch type；
If π/12>slope>- π/12, and 5 are not contained in type array, if putting bottom left section centered on s1 in addition to a left side Outside boundary and lower boundary, 0 degree of block number is spent and is less than more than -90, and point lower right-most portion is in addition to right margin and lower boundary centered on s2 Outside, the block number more than 0 degree and less than 90 degree；
If s1<2, and contain 1 in type array, it is judged to left dustpan type；
If s2<2, and contain 2 in type array, it is judged to right dustpan type；
If s1-s2>Thre1, s2<Thre2, and contain 2 in type array, it is judged to right dustpan type；
If s2-s1>Thre3, s1<Thre4, and contain 1 in type array, it is judged to left dustpan type；
If s1>Thre5, s2>Thre5, and contain 3 in type array, it is judged to bucket type；
Thre1, thre2, thre3, thre4, thre5 are threshold value with step d；
If slope >=π/12, and contain 1 in type array, it is judged to left dustpan type；
If slope≤- π/12, and contain 2 in type array, it is judged to right dustpan type；
G. it if being unsatisfactory for a, b, c, d, e, f, is judged to not judge.

Claims (1)

  1. A kind of 1. automatic fingerprint classification method, it is characterised in that the automatic fingerprint classification method includes training process and tested Journey, the specific steps are：
    (1) training is manually divided into six classes with fingerprint image, the fingerprint for being judged to first five class is trained.By the fingerprint image of H × W Block as being divided into 16 × 16 sizes, image share bi × bj blocks, calculation block pattern features, i.e. 4 direction block numbers, direction block 8 Direction position, by block directed graph feature and fingerprint pattern deposit feature array feature, 14 row, the 1st~4 row represent respectively： 1 block number of direction, 2 block number of direction, 3 block number of direction, 4 block number of direction, 8 direction positions of the 5th~13 row each piece of barycenter of storage, the 14th It is classified as fingerprint pattern；
    (2) block directed graph feature i.e. 4 direction block numbers and fingerprint pattern in feature array feature is taken, deletes and repeats to go, is deleted special Sign is that 4 direction block numbers are identical but different types of row, generates 4 direction block number database class_dire_num, 5 row, respectively 1 block number of direction, 2 block number of direction, 3 block number of direction, 4 block number of direction, fingerprint pattern；
    (3) feature array feature is taken, deletes the row that can classify by 4 direction block numbers, according to 8 direction position of direction block, is calculated 4 direction position of outgoing direction block, that is, quadrant deletes and repeats to go, delete feature i.e. 4 direction block numbers, 4 direction position of direction block it is identical but Different types of row, 4 direction location database class_dire4 of generation direction block, 14 row, the 1st~4 row represent respectively：Direction 1 Block number, 2 block number of direction, 3 block number of direction, 4 block number of direction, 4 direction positions of the 5th~13 row each piece of barycenter of storage, the 14th is classified as Fingerprint pattern；
    (4) feature array feature is taken, deletes the row that can classify by 4 direction block numbers, 4 direction position of direction block, deletes and repeats Row, deletes that feature i.e. 4 direction block numbers, 8 direction position of direction block be identical but different types of row, 8 direction position of generation direction block Database class_dire8,14 row, the 1st~4 row represent respectively：1 block number of direction, 2 block number of direction, 3 block number of direction, 4 pieces of direction Number, 8 direction positions of the 5th~13 row each piece of barycenter of storage, the 14th is classified as fingerprint pattern, feature i.e. 4 direction block numbers, direction block 8 Direction position is identical but the generation of different types of row can not judge database cantjudge, the same class_dire8 of form；
    (5) test fingerprint is pre-processed, calculation block pattern features i.e. 4 direction block numbers, 8 direction position of direction block；
    (6) 4 direction block number database class_dire_num are searched for 4 direction block numbers of test fingerprint, if it is found, output refers to Line type, test terminate, and otherwise perform step (7)；
    (7) according to 8 direction position of the direction block of test fingerprint, 4 direction position of outgoing direction block is calculated, with 4 direction block numbers and direction 4 direction position direction of search block of block, 4 direction location database class_dire4, if it is found, output fingerprint pattern, test knot Otherwise beam performs step (8)；
    (8) with 8 direction location database class_dire8 of 4 direction block numbers and 8 direction position direction of search block of direction block, if It finds, exports fingerprint pattern, test terminates, and otherwise performs step (9)；
    (9) database cantjudge can not be judged with 4 direction block numbers and 8 direction location finding of direction block, if found several Row, it would be possible in type deposit type array lx, otherwise remember lx (1)=0, represent this feature do not occur in training fingerprint Fingerprint；
    (10) singular point, block position and number where record singular point are found；
    (11) pseudo- singular point is deleted；
    (12) if not finding singular point, expanded scope searches for a central point；
    (13) judge fingerprint pattern by Henry sorting techniques, that is, singular point number, position and type array lx, what can not be judged sentences For that can not judge；
    The fingerprint is divided into six classes：Left dustpan type, right dustpan type, bucket type, arch form, cusped arch type and it can not judge；
    The calculating of block directed graph feature includes the following steps in the step (1)：
    A. block directed graph dirbo, section technique mask fmsk are calculated using gradient method；
    B. four direction feature is taken, block directed graph dirbo is changed to four directional diagram dir4 of block, method is
    C. directional diagram 1 is generated by four directional diagram dir4, with collar plate shape structural element corrosion position Fig. 1 that radius is 1, calculates corrosion The connection component of rear direction Fig. 1 calculates 8 direction position pos8 of the barycenter of connection component, is represented with 1-8, when connecting in component The number of pixel is less than or equal to ignore this block when 4；The block number and centroid position in direction 2,3,4 are similarly calculated, is stored in characteristic In group feature, each fingerprint a line；
    Block 4 direction position pos4 in direction is calculated according to 8 direction position pos8 of direction block in the step (3), is represented with 1~4, point Not Biao Shi 1~4 quadrant, method is
    Singular point includes central point and triangulation point in the step (10), includes the following steps：
    A. singular point central point number singularcore and triangulation point number singulardelta are initialized as 0；
    B. in fingerprint block directed graph dirbo except image border (the first row, last column, first row, last row) outside Each square calculates the variable quantity summation Poincare that direction rotates：
    △dk=dk- d(k+1)mod8K=0,1 ... 7
    If △ dk>=90, △ dk=△ dk- 180
    If △ dk≤ -90, △ dk=△ dk+180
    C. such as fruit block (i, j), block (i-1, j), block (i-1, j-1), the Poincare values of block (i, j-1) are 180, block (i, j) It is foreground blocks, and it is foreground blocks that its 3 × 3 neighborhood, which has 5 pieces or more, then central point number singularcore adds 1, by position i, j It is recorded in position array xx1 and yy1；
    D. such as fruit block (i, j), block (i-1, j) or block (i-1, j-2), block (i-1, j-1), the Poincare values of block (i, j-1) are equal It is -180, block (i, j) is foreground blocks, and it is foreground blocks that its 3 × 3 neighborhood, which has 5 pieces or more, triangulation point number singulardelta For 0 or singulardelta be 1 and distance is more than 3 between previous triangulation point, then triangulation point number singulardelta adds 1, position is recorded into i, j is in position array xx2 and yy2.
    It is as follows that pseudo- unusual point methods are deleted in the step (11)：
    A. to each block containing central point, if containing the block that Poincare values are -180, the block in its 5 × 5 field In without central point, central point number singularcore subtracts 1, the corresponding position recorded in delete position array xx1 and yy1；
    B. to each block containing triangulation point, if containing the block that Poincare values are 180 in its 5 × 5 field, in the block Without triangulation point, triangulation point number singulardelta subtracts 1, the corresponding position recorded in delete position array xx2 and yy2；
    The step (12) searches for a central point, and method is as follows：
    Be 180 and not in the block (i, j) of image border to Poincare values in (10) step result described in claim 1, i.e., the 1,2 row, 1,2 rows reciprocal, the 1st, 2 row, inverse 1,2 arrange, the variable quantity summation Poincare rotated by 5 × 5 neighborhoods calculating direction：
    △dk=dk-d(k+1)mod8K=0,1 ... 7
    If △ dk>=90, △ dk=△ dk-180
    If △ dk≤ -90, △ dk=△ dk+180
    If Poincare (i, j)=180, central point number singularcore=1, by position i, j is recorded in position array In xx1 and yy1, step (13) is performed；
    The sorting technique of the step (13) is as follows：
    A. if central point number singularcore=2 or triangulation point number singulardelta=2, and type array lx In containing 3 or lx (1)=0, be judged to bucket type；
    B. if central point number singularcore=1 or triangulation point number singulardelta=1, type array have two A element, and one of them is 4, then is judged to another type；
    C. if central point number singularcore=0 and triangulation point number singulardelta=0, and in type array Containing element 4, it is judged to arch form；
    D. if central point number singularcore=1 and triangulation point number singulardelta=0, sentence by the following method It is disconnected：
    It is not arch form, but can not judge if central point bottom left section or lower right-most portion prospect are less than 5 pieces；
    If bottom left section is put centered on s1 other than left margin and lower boundary, the block number more than -90 degree and less than 0 degree, during s2 is Heart point lower right-most portion is other than right margin and lower boundary, the block number more than 0 degree and less than 90 degree；
    If s1<2, and contain 1 in type array, it is judged to left dustpan type；
    If s2<2, and contain 2 in type array, it is judged to right dustpan type；
    If s1-s2>Thre1, s2<Thre2, and contain 2 in type array, it is judged to right dustpan type；
    If s2-s1>Thre3, s1<Thre4, and contain 1 in type array, it is judged to left dustpan type；
    If s1>Thre5, s2>Thre5, and contain 3 in type array, it is judged to bucket type；
    Thre1, thre2, thre3, thre4, thre5 are threshold value；
    E. if central point number singularcore=0 and triangulation point number singulardelta=1, triangulation point position is calculated Put relative position yy2 (1)/bj (bj be picture traverse, units chunk) in figure；
    If yy2/bj>Thre6, and contain 1 in type array, it is judged to left dustpan type；
    If yy2/bj<Thre7, and contain 2 in type array, it is judged to right dustpan type；
    Thre6, thre7 are threshold value；
    F. if central point number singularcore=1 and triangulation point number singulardelta=1, calculate central point with Triangulation point line and center line angleJudge by the following method：
    If π/12>slope>- π/12, and contain 5 in type array, it is judged to cusped arch type；
    If π/12>slope>- π/12, and 5 are not contained in type array, if put centered on s1 bottom left section in addition to left margin and Outside lower boundary, 0 degree of block number is spent and is less than more than -90, and point lower right-most portion is more than other than right margin and lower boundary centered on s2 0 degree and the block number less than 90 degree；
    If s1<2, and contain 1 in type array, it is judged to left dustpan type；
    If s2<2, and contain 2 in type array, it is judged to right dustpan type；
    If s1-s2>Thre1, s2<Thre2, and contain 2 in type array, it is judged to right dustpan type；
    If s2-s1>Thre3, s1<Thre4, and contain 1 in type array, it is judged to left dustpan type；
    If s1>Thre5, s2>Thre5, and contain 3 in type array, it is judged to bucket type；
    Thre1, thre2, thre3, thre4, thre5 are threshold value with step d；
    If slope >=π/12, and contain 1 in type array, it is judged to left dustpan type；
    If slope≤- π/12, and contain 2 in type array, it is judged to right dustpan type；
    G. it if being unsatisfactory for a, b, c, d, e, f, is judged to not judge.
CN201711490895.1A 2017-12-30 2017-12-30 Automatic fingerprint classification method Active CN108268836B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711490895.1A CN108268836B (en) 2017-12-30 2017-12-30 Automatic fingerprint classification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711490895.1A CN108268836B (en) 2017-12-30 2017-12-30 Automatic fingerprint classification method

Publications (2)

Publication Number Publication Date
CN108268836A true CN108268836A (en) 2018-07-10
CN108268836B CN108268836B (en) 2021-04-09

Family

ID=62772963

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711490895.1A Active CN108268836B (en) 2017-12-30 2017-12-30 Automatic fingerprint classification method

Country Status (1)

Country Link
CN (1) CN108268836B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112199049A (en) * 2020-10-22 2021-01-08 Tcl通讯（宁波）有限公司 Fingerprint storage method and device and terminal

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030002719A1 (en) * 2001-06-27 2003-01-02 Laurence Hamid Swipe imager with multiple sensing arrays
CN1595425A (en) * 2004-07-13 2005-03-16 清华大学 Method for identifying multi-characteristic of fingerprint
CN102368241A (en) * 2011-09-07 2012-03-07 常州蓝城信息科技有限公司 Multi-stage fingerprint database search method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030002719A1 (en) * 2001-06-27 2003-01-02 Laurence Hamid Swipe imager with multiple sensing arrays
CN1595425A (en) * 2004-07-13 2005-03-16 清华大学 Method for identifying multi-characteristic of fingerprint
CN102368241A (en) * 2011-09-07 2012-03-07 常州蓝城信息科技有限公司 Multi-stage fingerprint database search method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112199049A (en) * 2020-10-22 2021-01-08 Tcl通讯（宁波）有限公司 Fingerprint storage method and device and terminal
CN112199049B (en) * 2020-10-22 2023-10-20 Tcl通讯（宁波）有限公司 Fingerprint storage method, fingerprint storage device and terminal

Also Published As

Publication number Publication date
CN108268836B (en) 2021-04-09

Similar Documents

Publication Publication Date Title
Rosin Training cellular automata for image processing
US6445832B1 (en) Balanced template tracker for tracking an object image sequence
CN108470354A (en) Video target tracking method, device and realization device
US8036468B2 (en) Invariant visual scene and object recognition
CN111844101B (en) Multi-finger dexterous hand sorting planning method
EP3813661A1 (en) Human pose analysis system and method
CN105678737B (en) A kind of digital picture angular-point detection method based on Radon transformation
CN107564048A (en) Based on bifurcation feature registration method
Shneier Using pyramids to define local thresholds for blob detection
CN112364865B (en) Method for detecting small moving target in complex scene
CN111951283A (en) Medical image identification method and system based on deep learning
Rout et al. Walsh–Hadamard-kernel-based features in particle filter framework for underwater object tracking
CN111368865B (en) Remote sensing image oil storage tank detection method and device, readable storage medium and equipment
CN106033613B (en) Method for tracking target and device
CN108268125A (en) A kind of motion gesture detection and tracking based on computer vision
CN108268836A (en) Automatic fingerprint classification method
Khan et al. Segmentation of single and overlapping leaves by extracting appropriate contours
Yang et al. Method for building recognition from FLIR images
Biswas Finger detection for hand gesture recognition using circular hough transform
JP7323849B2 (en) Division recognition method, division recognition device and program
CN114550062A (en) Method and device for determining moving object in image, electronic equipment and storage medium
CN108038864B (en) Method and system for extracting animal target image
Torsello et al. A skeletal measure of 2D shape similarity
Połap et al. Detection of important features from images using heuristic approach
Vu et al. Automatic improvement of graph based image segmentation

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