CN114120378A - Three-level classification fingerprint identification method - Google Patents
Three-level classification fingerprint identification method Download PDFInfo
- Publication number
- CN114120378A CN114120378A CN202111477702.5A CN202111477702A CN114120378A CN 114120378 A CN114120378 A CN 114120378A CN 202111477702 A CN202111477702 A CN 202111477702A CN 114120378 A CN114120378 A CN 114120378A
- Authority
- CN
- China
- Prior art keywords
- fingerprint
- points
- point
- image
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000000605 extraction Methods 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims description 14
- 238000010586 diagram Methods 0.000 claims description 13
- 238000013441 quality evaluation Methods 0.000 claims description 8
- 238000003709 image segmentation Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 2
- 238000007670 refining Methods 0.000 claims description 2
- 230000011218 segmentation Effects 0.000 claims description 2
- 238000013500 data storage Methods 0.000 claims 2
- 101100074336 Xenopus laevis ripply2.1 gene Proteins 0.000 claims 1
- 238000009499 grossing Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 7
- 238000011160 research Methods 0.000 abstract description 6
- 238000007635 classification algorithm Methods 0.000 abstract description 4
- 239000000284 extract Substances 0.000 abstract description 3
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000001303 quality assessment method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012790 confirmation Methods 0.000 description 2
- 238000003825 pressing Methods 0.000 description 2
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 210000004243 sweat Anatomy 0.000 description 1
Images
Landscapes
- Collating Specific Patterns (AREA)
Abstract
The invention relates to the technical field of artificial intelligence identification, in particular to a fingerprint identification method. The fingerprint feature extraction part comprises minutiae extraction and singular point extraction, and the fingerprint identification technology mainly extracts feature information from the combination of ridges and valleys by analyzing the global features of the fingerprint and the local features of the fingerprint, wherein the local feature information is used in a fingerprint matching stage; fingerprint classification is commonly found in identification systems with large-capacity fingerprint databases, which divides the fingerprint database into a plurality of subclasses, so as to match the fingerprint to be identified with the fingerprint of the same class in the database; fingerprint matching is to judge whether two given fingerprint feature information come from the same finger by measuring the similarity of the two given fingerprint feature information. The classification method in the fingerprint identification system is taken as a main research direction, a three-level classification algorithm of the fingerprint and a fingerprint identification system based on classification characteristics are researched and realized, and the whole fingerprint database is reduced into a homologous fingerprint set similar to the fingerprint to be identified.
Description
Technical Field
The invention relates to the technical field of artificial intelligence identification, in particular to a fingerprint identification method.
Background
The realization of personal identity authentication by using physiological characteristics or behavioral characteristics of a human body is a hot point of research. Fingerprint identification technology is the most widely used one of biometric identification technologies due to its uniqueness and stability. With the advent of the digital age, biometric identification technology will gradually replace traditional authentication methods such as keys and passwords.
A fingerprint identification method and fingerprint identification system, application (patent) No.: CN200810065665.5, application date: 2008.01.30. A fingerprint identification method is provided, which comprises the following steps: collecting a fingerprint image, extracting fingerprint characteristics of the fingerprint image, and performing characteristic matching on the fingerprint characteristics and a fingerprint template; and after matching is successful, updating the characteristics of the fingerprint template according to the fingerprint characteristics. The invention also provides a fingerprint identification system. The technical scheme of the invention has the following beneficial effects: the latest fingerprint template is subjected to feature updating according to the fingerprint features successfully matched, the latest fingerprint feature information in the fingerprint peeling process can be mastered at any time, and the accurate matching of the fingerprints in each stage in the non-peeling and peeling processes can be realized by matching with the original fingerprint template.
A quick fingerprint identification method based on a singular topological structure applies for (patent) No: CN200610112936.9, application date: 2006.09.13. a quick fingerprint identification method based on singular point topological structure is disclosed, which can quickly process and identify the fingerprint image in automatic fingerprint identification system by using the topological structure of singular points in the fingerprint. The method classifies the fingerprint database according to the classification characteristic and the position stability of the singular points in the fingerprint image, selects the most reliable singular points as reference points, constructs the topological structure of the singular points of the fingerprint and constructs a template for identification. The method comprises the steps of firstly detecting true singular points and false singular points in a fingerprint image, classifying a database according to the relative positions of the singular points, generating a topological structure adjacent to the singular points, using the topological structure for constructing a fingerprint template, and accelerating the comparison process of fingerprints.
The invention optimizes the classification method in the prior art, researches and realizes a three-level classification algorithm of the fingerprint and a fingerprint identification system based on classification characteristics, and reduces the whole fingerprint database into a homologous fingerprint set similar to the fingerprint to be identified.
Disclosure of Invention
The invention solves the defects of the prior art and adopts the following technical scheme: a fingerprint identification system and a using method thereof comprise the following steps: collecting and storing information; preprocessing an image; extracting fingerprint features; classifying the fingerprints; matching fingerprints; storing data; the metrics identify several parts. The image preprocessing mainly comprises the processes of image segmentation, quality evaluation, binarization, thinning and the like, and the fingerprint image is preprocessed to enhance the fingerprint ridge-valley contrast and reduce pseudo characteristic information; the fingerprint feature extraction part comprises minutiae extraction and singular point extraction, and the fingerprint identification technology mainly extracts feature information from the combination of ridges and valleys by analyzing the global features of the fingerprint and the local features of the fingerprint, wherein the local feature information is used in a fingerprint matching stage; fingerprint classification is commonly found in identification systems with large-capacity fingerprint databases, which divides the fingerprint database into a plurality of subclasses, so as to match the fingerprint to be identified with the fingerprint of the same class in the database; the fingerprint matching judges whether the two given fingerprint feature information come from the same finger or not by measuring the similarity of the two given fingerprint feature information;
the invention has the beneficial effects that: because the population base number is huge, the research of the fingerprint identification technology gradually turns to the automatic fingerprint identification based on a large-capacity fingerprint database, in order to find a fingerprint template matched with the fingerprint to be identified, an effective index needs to be established in the fingerprint database, otherwise, the database is subjected to full-table scanning, the time cost is high, and the requirement of people on a high-efficiency system is not met. Therefore, the classification method in the fingerprint identification system is taken as a main research direction, a three-level fingerprint classification algorithm and a fingerprint identification system based on classification characteristics are researched and realized, and the whole fingerprint library is reduced into a homologous fingerprint set similar to the fingerprint to be identified.
Drawings
FIG. 1 is a schematic diagram of a technical route of an automatic fingerprint identification system;
FIG. 2 is a diagram of fingerprint feature points;
FIG. 3 is a schematic view of different fingerprint shapes;
FIG. 4 is a schematic diagram of a pattern area, core points, and triangle points;
FIG. 5 is a diagram of a pretreatment process;
FIG. 6 is a schematic diagram of an eight neighborhood;
FIG. 7 is a schematic diagram of three-level classification of fingerprints;
FIG. 8 is a diagram of three-level classification of fingerprints.
Detailed Description
A fingerprint recognition system designed herein as shown in fig. 1 mainly includes learning and recognition, and stores the learned fingerprints through image acquisition, image preprocessing, fingerprint quality evaluation, feature extraction, fingerprint classification, etc., as the basis for later recognition.
Introduction of automatic fingerprint identification system
Fingerprinting techniques generally comprise the following stages: the preprocessing mainly comprises the processes of image segmentation, quality evaluation, binarization, thinning and the like, and the fingerprint image can enhance the fingerprint ridge-valley contrast and reduce the pseudo characteristic information through preprocessing; the characteristic extraction part comprises minutiae extraction and singular point extraction, and the fingerprint identification technology mainly extracts characteristic information from the combination of ridges and valleys by analyzing the global characteristic of the fingerprint and the local characteristic of the fingerprint, wherein the local characteristic information is used in a fingerprint matching stage; fingerprint classification is commonly found in identification systems with large-capacity fingerprint databases, and aims to stably and reliably divide the fingerprint database into a plurality of subclasses so as to match the fingerprint to be identified with the fingerprint of the same class in the database; fingerprint matching is to judge whether two given fingerprint feature information come from the same finger by measuring the similarity of the two given fingerprint feature information.
Fingerprint basic characteristics
The fingerprint pattern is not continuous, smooth and straight, but is often interrupted, branched or turned. These break points, bifurcation points and inflection points are referred to as "feature points". The feature points provide confirmation of fingerprint uniqueness, most typically intersections, cores, ridge tails, islands, triangles, sweat pores, etc., as shown in fig. 2.
General features
The overall fingerprint features are those that can be directly observed by human eyes. Including the pattern, the pattern area, the core point, the triangle point, the number of the patterns, etc. Fingerprints are generally classified into three categories, i.e., circular, arched and spiral according to the direction and distribution of ridges, as shown in fig. 3.
The pattern area, i.e. the area of the fingerprint that includes the global features, from which it can be distinguished which type of fingerprint belongs to, as shown in fig. 4.
The core point is located at the progressive center of the fingerprint lines and serves as a reference point when the fingerprint is read and compared. Many algorithms are based on core points, i.e. only fingerprints with core points can be processed and identified.
The triangle point is located at the first bifurcation or breakpoint from the core point, or at the convergence, solitary point, turning point of the two lines, or points to these singular points. The triangular dots provide the beginning of the count trace of the fingerprint line.
The number of lines, i.e. the number of fingerprint lines in the pattern area. When calculating the fingerprint lines, the core point and the triangle point are connected first, and the number of the intersection of the connecting line and the fingerprint lines can be regarded as the number of the fingerprint lines.
Local features
The characteristics of the local characteristic fingerprint nodes. The fingerprint is not continuous, smooth and straight, and often has branches, folds or breaks. These intersections, inflection points, or breakpoints are called "feature points" which provide confirmation of fingerprint uniqueness, and the main parameters of the feature points include:
the direction is as follows: the direction in which the feature points are located relative to the core points.
Curvature: the speed at which the direction of the grain changes.
Position: the position coordinates of the nodes are described by x/y coordinates. It may be absolute coordinates or relative coordinates to triangular points (or feature points).
(II) fingerprint preprocessing
The fingerprint preprocessing is a precondition of fingerprint identification, and can effectively solve the problems that the acquired fingerprint is an image containing various noises due to the possible existence of the problems of the degree of pressure, the degree of dryness and wetness of the finger, the performance of the acquisition device and the like in the fingerprint acquisition process, and the success or failure of the fingerprint identification is directly influenced by the quality of the fingerprint preprocessing, and the processing process is as shown in the following figure 5.
(III) fingerprint feature extraction
Feature extraction is to extract information representing features of a fingerprint image. The method for extracting the features from the refined binary image is simple, and only a 3 x 3 template is needed to extract the end points and the bifurcation points.
(IV) fingerprint matching
Fingerprint matching is the process of comparing two fingerprints to determine if they are homologous, i.e., if the two fingerprints originate from the same finger. The method adopts matching based on detail feature points, and comprises two steps of reference point positioning and accurate matching, wherein the purpose of the reference point positioning is to align two feature point sets, the purpose of the reference point positioning is to accurately match the two feature point sets, and the purpose of the accurate matching is to obtain the conclusion whether the two feature point sets are from the same finger or not by recording the number of similar feature point pairs.
Secondly, preprocessing of fingerprints and feature extraction
In the fingerprint acquisition process, a sensor is subjected to digital processing to form a digital image, the acquired fingerprint image has insufficient characteristic information due to the fact that the finger may have molting, moisture, unstable performance of the sensor and the like, and the fingerprint needs to undergo preprocessing and characteristic extraction in a system in order to realize correct classification and correct matching of the fingerprint image.
A directional field and a frequency field
The direction information of the fingerprint can reflect the basic pattern of the fingerprint image, and can be divided into a vector distribution diagram (point direction) based on each pixel point and a direction vector distribution diagram (block direction) based on each block according to the degree of fineness. The point directional diagram is formed by the real directions of the pixel points, and the block directional diagram is the gradient minimum direction of the gray level change in the neighborhood of w x w by taking the pixel points as the center. The ridge frequency of a fingerprint refers to the number of ridges contained in a unit distance, and is a very important feature of a fingerprint image. The method adopts a traditional gradient method to realize the estimation of the fingerprint ridge direction field, and adopts the following method to estimate the ridge frequency, and comprises the following specific steps:
(1) dividing the image into a plurality of sub-blocks with the size w (16 w 16); and defining pixel point (i, j) as center and length and width aslWindow of direction of wl=32) The x-axis is perpendicular to the stripe line direction of the block, and the y-axis is perpendicular to the x-axis;
(2) traversing pixel points in the sub-blocks, summing the gray values in each sub-block, and dividing the sum by the image block w to obtain a gray discrete signal X (m) in the horizontal direction in the direction window, wherein m is any value between 0 and 1.
(3) The average period is calculated. Searching an extreme value in X, recording the position and the number of the extreme value, calculating the mean period T (i, j) of the ridge line, if two continuous peak values do not exist in X, determining that the block contains a singular point or a minutia point, and enabling the frequency value to be equal to 0;
(4) the average frequency is calculated. If T (i, j) is equal to [3,25], (i, j) is the image coordinate system, the average frequency is the reciprocal of the average period, then the frequency value F (i, j) =1/T (i, j) of the sub-block, and the effective frequency value is in the range of [0.04,0.33], if the value of F (i, j) is not in the effective range, in the same step (3), the block frequency value is equal to 0.
(II) fingerprint image segmentation
There are two important areas of the fingerprint image that need attention, namely the region of interest and the region without intensity variation at the edge of the image. The method adopts a frequency field-based segmentation method, and processes the situation of holes and black spots existing on the boundary, and comprises the following specific steps:
(1) defining the image of the region of Interest as Interest, traversing the ridge line frequency field, setting the gray value of Interest at (i, j) to be 255 if the frequency value is between [0.04,0.33], otherwise setting the gray value to be 0;
(2) the template B is a 16-system value 0x80 set at four positions in the 8-neighborhood graph in the vertical direction and the horizontal direction; translating the template B on Interest, and traversing each pixel point in the graph;
(3) and if the pixel gray value of the central position of the template B is 255, carrying out 'OR' operation on the Interest and the template B. If the pixel gray value after traversing is not 0, the pixel gray value is equal to 255;
(4) and if the pixel gray value of the central position of the template B is 0, performing AND operation on Interest and the template B, and if the pixel gray value after traversal is not 255, making the pixel gray value equal to 0.
(III) fingerprint image quality assessment
The quality of the fingerprint image is mainly influenced by factors such as the pressing position of the finger, the pressing pressure degree, the finger dryness and wetness degree, the sensor performance and the like, and the fingerprint image quality evaluation flow is integrated into the preprocessing to screen the quality of the fingerprint image. A feature-based quality assessment method is used herein to reduce the system false positive and false negative rates, which can effectively filter low quality fingerprints and give relevant prompts for the user to re-acquire fingerprint images that can be used for subsequent classification matching. The quality assessment method herein is as follows:
(1) and evaluating the quality of the effective area of the image. The preset percentage T1 is 30%, and the number of pixel points contained in the white area can be quickly counted by combining with the fingerprint image segmentation part. The effective area ratio is the area of the region of interest divided by the total area. When R < T1, quality is not suitable; otherwise, the second stage evaluation may proceed.
(2) And evaluating the gray contrast of the effective area. Presetting a lower limit T2 of gray scale contrast to be 0.4, calculating a gray scale mean value and a gray scale variance of an effective area of each fingerprint image pixel point, and dividing the gray scale mean value by the gray scale variance value to be used as a gray scale contrast C; and when the gray contrast C meets the lower limit T2, performing quality evaluation of the third stage, and prompting the user to re-acquire the fingerprint for the fingerprint which does not meet the condition.
(3) And evaluating the centroid offset. But the geometric center of the active area is offset by less than 1/2 for both the length and height of the image of interest of the fingerprint, the image can be evaluated by the third stage, otherwise the rejection of the fingerprint prompts the user to re-acquire.
(IV) binarization and refinement of fingerprint image
1. Binarization method
Because the ridge line in the fingerprint image is darker than the valley line by the imaging principle, the purpose of binarization is to highlight the ridge line, and then the valley line part is divided into a background area, the ridge line is processed into pixel points with the gray value of 255, and the valley line part is processed into pixel points with the gray value of 0, so that a good data source is provided for the subsequent algorithm.
Setting I (I, j) to describe a gray value with a coordinate point (I, j) in the fingerprint image, wherein I belongs to [0, h-1], j belongs to [0, w-1], and h w describes the size of the image; and (3) setting a probability function P (x) to describe the probability of the gray value x appearing in the image, wherein x belongs to [0,255], selecting the intermediate value 128 as the value of the threshold value T, and enabling the gray value of the pixel point larger than T to be equal to 255 and the gray value of the pixel point smaller than T to be equal to 0.
2. Refining
For ease of algorithm description, an eight neighborhood model is defined herein, as shown in FIG. 6. Namely, the current point is taken as the center, the current point and eight points adjacent to the center point form a 3 x 3 template, the position relationship between each adjacent point and the center point forms an eight-neighborhood model, A represents the current center point, and P1-P8 represent adjacent points in 8 directions of the center point respectively.
The method comprises the following specific steps: performing horizontal traversal on the binary image once, and if the gray values of P3 and P7 are the same as the gray value of A, keeping the gray value of the point A unchanged; otherwise, if the value is 0, the gray value of the point A is not changed, and if the value is 1, the gray value of the point A is set to be 0; a vertical traversal of the image is made until there is no change in the gray value of the pixel.
(V) positioning and extracting singular points of fingerprint
The method for extracting singularities used herein comprises the following steps:
(1) taking an 8-neighborhood graph of the pixel point as an example, making a closed curve surrounding the pixel point, selecting a certain point on the curve as a starting point P0 of the traversal pixel point, selecting the next pixel point P1 in a clockwise direction, and so on until the next pixel point returns to P0 again;
(2) taking the block direction values of the sub-blocks respectively taking two adjacent pixel points as the centers to make a difference, and recording the difference as delta (k);
(3) if-pi/2 < | delta (k) | < pi/2, then the same is maintained; if | Δ (k) | < = -pi/2, Δ (k) = Δ (k) + pi; otherwise, Δ (k) = Δ (k) -pi; next, calculating the sum of the direction differences on the closed curve;
(4) if it isIf the value is not less than pi, the surrounding curve contains a core point; if it isIf the value is = pi, the surrounding curve contains triangular points; storing the information of the singular points, such as coordinates, directions and types into a feature point set; if the two calculation results are not met, otherwise, no singular point exists in the region.
Fingerprint identification research based on classification characteristics
For a system of a large-capacity fingerprint database, a final conclusion can be obtained at the expense of time by sequentially matching a target fingerprint with template fingerprints in the database, the system is crashed seriously, if the template fingerprints are stored in a classified manner according to the characteristics of the template fingerprints, when one fingerprint is input, the target fingerprint can be processed and judged to be classified, and only one fingerprint needs to be matched in homologous subclasses, so that the searching time is shortened, and the matching times are reduced. In the process of using multi-stage classification, the problem of deviation of classification is particularly required to be concerned, and the stability and consistency of fingerprint patterns are strong, so the patterns are used as first-stage classification features; secondly, the stability of the number of ridges between singular points is stronger than the average period of the ridges, and the former is used as the second-stage classification feature, and the latter is used as the third-stage classification feature, as shown in fig. 7.
(one) one-level classification based on pattern type
The fingerprint pattern can be determined by the number and relative position of singular points, and the specific pattern can be judged according to the included angle between the connecting line between the two points and the reference coordinate axis of the fingerprint image, and the specific method comprises the following steps:
(1) connecting the central point and the triangular point;
(2) when the number of the core points is more than or equal to 2, the core points are spiral lines;
(3) when the number of core points is less than 1, the shape is a bow shape;
(4) when the number of the core points is equal to 1 and the number of the triangle points is not equal to 1, the core points are in the miscellaneous class;
(5) when the number of the core points is equal to 1 and the number of the triangle points is also 1, calculating the direction difference theta between the line segment direction of the connecting line and the direction of the central point; if the Δ is more than pi/9, the shape is the right skip; if the Δ is less than-pi/9, the left skip is a left skip; if the | Δ | is less than or equal to Pi/9, the book belongs to the book.
Two-stage classification of number of singularity-based inter-ridges
The second-stage classification mainly reduces the data quantity difference in each category, so the second-stage classification is carried out on the left skip shape, the right skip shape and the spiral shape. The calculation of the ridge number is based on the connection line between singular points, and the number of the intersection points of the straight line and the ridge is calculated by adopting the following method:
(1) traversing 8 neighborhoods of points on the connecting line along the connecting line direction by taking the triangular points or the core points as starting points, judging the number of ridge line pixel points, and if the number is less than 2, not calculating to be an intersection point; if the number is 2, executing the step (2); if the number is 3, executing the step (3);
(2) calculating the distance s between two ridge line pixel points, and if s =1, not calculating an intersection point; otherwise, the total number of the intersection points is added by 1;
(3) counting the number of the pixel points containing 3 ridge lines, and if the number is an even number, adding 1 to the total intersection point number;
(4) traversing the intersection point set, calculating the difference value of the horizontal and vertical coordinates of two adjacent intersection points, if one of the two difference values is equal to 1, selecting one of the two intersection points to be removed from the intersection point set, and subtracting one from the total number of the intersection points.
The ridge number is herein divided into 13 classes Ri (1 < = i < = 13). Setting the upper limit of the search radius as r, calculating the ridge number N between singular points of the fingerprint to be identified, and describing the method for searching the homologous fingerprint in the database as follows:
(1) obtaining a corresponding category Ri according to the ridge line number N, and setting a search radius c = 0;
(2) searching fingerprints in the corresponding categories and sequentially comparing the fingerprints with the fingerprints to be identified, and finishing matching if the matching is successful; continuing the following steps if the unmatching is successful;
(3) gradually increasing the search radius by taking 1 as a gradient, repeating the steps (4) and (5) until the matching is finished when the search radius is larger than the upper limit of the search radius, and declaring the search to fail;
(4) when the sum of the category and the search radius is less than 14, the data are one by one at Ri+cComparing the fingerprint template in the category with the fingerprint to be identified, and if the matching is successful, ending the searching process; otherwise, executing the step (5);
(5) when the category is larger than the search radius, the category is one by one in Ri-cComparing the fingerprint template in the category with the fingerprint to be identified, and if the matching is successful, ending the search; otherwise, the matching fails.
(III) three-stage classification based on ridge line average period
The third-level classification is carried out only by using the continuity characteristic, namely the ridge line average period of the fingerprint center area, of all fingerprints subjected to the second-level classification, and the existing classes are further refined. Firstly, a 7-by-7 mean filter is constructed to carry out convolution operation on the image frequency value, and partial pixel points with the frequency value of 0 are corrected into pixel points with the frequency value, namely, the image frequency is smoothed, so that a more accurate frequency graph can be obtained. After the frequency value is obtained, the fingerprint image is sorted into 60 × 80 images by taking the geometric center of the effective area of the fingerprint as a standard, and the average period value of the ridge line can be obtained by calculating the frequency sum of each image and then dividing the frequency sum by the size of the whole image (60 × 80). The average period is first associated with the number of fingerprints, and then the ridge average period is divided into 10 classes.
The three classification methods are adopted to show the classification process of the system, and the specific flow is shown in fig. 8. By carrying out three-stage classification processing on the fingerprints, the matching times with the fingerprints to be identified are reduced, subclasses can be effectively divided, and the matching time is reduced.
A fingerprint identification method based on three-level classification features is characterized in that information such as the number and direction of singular points is utilized in the first-level classification, and fingerprints are divided into six categories such as bow, account bow, left skip, right skip, spiral and impurity which are difficult to classify; and performing third-level classification on the result of the second-level classification according to the distribution condition of the ridge line average period at the middle position of the fingerprint mode area, and subdividing 10 subclasses into second-level subclasses.
According to the statistical data of the predecessors, the 6 major classes related to the first-level fingerprint classification are not uniformly classified, and the corresponding classification accounts for the following table 1:
the secondary fingerprints are classified in 13 types, and the corresponding proportion is as shown in table 2:
the three-level 10 classes of fingerprints are classified according to the proportion shown in Table 3
The text uses entropy to explain the classification efficiency of the text, and the classification efficiency function based on the entropy is shown as follows
Wherein the content of the first and second substances,for the number of classification features, k is the number of classes, and p is the probability that any fingerprint belongs to a class. According to the classification efficiency function of the information entropy, the three-level classification efficiency can be obtained, and the probability data required in the first-level classification can be known from the table 1
From Table 2, it can be seen that
As can be seen from Table 3
The classification efficiency of the multi-stage classification algorithm herein is therefore 3.52X9.75X8.14=279.36, and theoretically the bow and the miscellaneous fingerprints of the first stage classification are removed and the corresponding fingerprints are classified as
(6-2)X 13 X 10=520
The relative classification efficiency is 279.36/520=0.5372, and the relative classification efficiency of the traditional fingerprint identification is 0.3113, so that the relative classification efficiency of the text is high.
Claims (6)
1. A three-level classification fingerprint identification method is characterized by comprising the following steps: the method comprises the following steps of information acquisition and storage, image preprocessing, fingerprint feature extraction, fingerprint classification, fingerprint matching, data storage and index identification, wherein the image preprocessing comprises image segmentation, quality evaluation, binaryzation and refinement; the fingerprint feature extraction comprises minutiae extraction and singular point extraction, and the fingerprint identification is realized by analyzing the global features of the fingerprint and the local features of the fingerprint, analyzing the combination of ridges and valleys and extracting feature information from the combination, wherein the local feature information is used for fingerprint matching; the data storage is carried out through a fingerprint database, the stored data of the fingerprint database are divided into a plurality of subclasses, and the fingerprints to be identified are matched with the fingerprints of the same class in the database; fingerprint matching judges whether fingerprint characteristic information comes from the same finger or not by measuring the similarity of the fingerprint characteristic information.
2. The three-level classification fingerprint identification method according to claim 1, wherein the preprocessing and feature extraction of the fingerprint are as follows:
a directional field and a frequency field
The direction information of the fingerprint can reflect the basic pattern of the fingerprint image, and can be divided into a vector distribution pattern point direction based on each pixel point and a direction vector distribution pattern block direction based on each block according to the fineness degree, wherein a point directional diagram is formed by the real directions of the pixel points, a block directional diagram is in the gradient minimum direction of gray level change in a w x w neighborhood by taking the pixel points as the center, the ridge line frequency of the fingerprint refers to the ridge line number contained in a unit distance, the estimation of a fingerprint ridge line direction field is realized by adopting a gradient method, and the ridge line frequency is estimated by adopting the following method, and the specific steps are as follows:
(1) dividing the image into a plurality of sub-blocks with the size w (16 w 16); and defining pixel point (i, j) as center and length and width aslWindow of direction of wl=32) The x-axis is perpendicular to the stripe line direction of the block, and the y-axis is perpendicular to the x-axis;
(2) traversing pixel points in the sub-blocks, summing the gray values in each sub-block, and dividing the sum by the image express delivery w to obtain a gray discrete signal X (m) in the horizontal direction in a direction window, wherein m is any value between 0 and 1;
(3) calculating an average period, searching for an extreme value in X, recording the position and the number of the extreme value, calculating a ridge line average period T (i, j), if two continuous peak values do not exist in X, determining that the block contains a singular point or a minutia point, and enabling the frequency value to be equal to 0;
(4) calculating an average frequency, wherein T (i, j) belongs to [3,25], (i, j) is an image coordinate system, the average frequency is the reciprocal of an average period, the frequency value F (i, j) =1/T (i, j) of the sub-block is in the range of [0.04,0.33], and if the value of F (i, j) is not in the effective range, the same step (3) is carried out, and the block frequency value is equal to 0;
(II) fingerprint image segmentation
The fingerprint image key area is respectively an interesting area and an image edge non-gray degree change area, a frequency field-based segmentation method is adopted, and the conditions of holes and black points existing on the boundary are processed, and the specific steps are as follows:
(1) defining the image of the region of Interest as Interest, traversing the ridge line frequency field, setting the gray value of Interest at (i, j) to be 255 if the frequency value is between [0.04,0.33], otherwise setting the gray value to be 0;
(2) the template B is a 16-system value 0x80 set at four positions in the 8-neighborhood graph in the vertical direction and the horizontal direction; translating the template B on Interest, and traversing each pixel point in the graph;
(3) if the pixel gray value of the central position of the template B is 255, carrying out 'OR' operation on the Interest and the template B;
if the pixel gray value after traversing is not 0, the pixel gray value is equal to 255;
(4) and if the pixel gray value of the central position of the template B is 0, performing AND operation on Interest and the template B, and if the pixel gray value after traversal is not 255, making the pixel gray value equal to 0.
3. The three-level classification fingerprint identification method according to claim 2, characterized in that the fingerprint image quality evaluation:
(1) evaluating the quality of the effective area of the image, presetting the percentage T1 to be 30%, combining with a fingerprint image segmentation part, counting the number of pixel points contained in a white area, wherein the effective area ratio is the area of the interested area divided by the total area, and when R is less than T1, the quality is not appropriate; otherwise, continuing to perform the second-stage evaluation;
(2) evaluating the gray scale contrast of the effective area, presetting a lower limit T2 of the gray scale contrast to be 0.4, calculating the gray scale mean value and the gray scale variance of the effective area of each fingerprint image pixel point, and dividing the gray scale mean value by the gray scale variance value to be used as the gray scale contrast C; when the gray contrast C meets the lower limit T2, performing quality evaluation in the third stage, and prompting the user to re-collect the fingerprint for the fingerprint which does not meet the condition;
(3) and evaluating the centroid offset, and if the geometric center offset of the effective area is smaller than 1/2 in both length and height of the interested image of the fingerprint, evaluating through the third stage, and otherwise rejecting the fingerprint to prompt the user to re-acquire the fingerprint.
4. The three-level classification fingerprint identification method according to claim 3, characterized in that the binarization and refinement of the fingerprint image:
(1) binarization method
Then dividing the valley line part into background area, processing the ridge line into pixel points with the gray value of 255, processing the valley line part into pixel points with the gray value of 0,
setting I (I, j) to describe a gray value with a coordinate point (I, j) in the fingerprint image, wherein I belongs to [0, h-1], j belongs to [0, w-1], and h w describes the size of the image; setting a probability function P (x) to describe the probability of the gray value x appearing in the image, wherein x belongs to [0,255], selecting an intermediate value 128 as a threshold value T, and enabling the gray value of a pixel point larger than T to be equal to 255 and the gray value of a pixel point smaller than T to be equal to 0;
(2) refining
Defining an eight-neighborhood model, namely, taking the current point as the center, forming a 3 x 3 template with eight points adjacent to the center point, forming the eight-neighborhood model by the position relationship between each adjacent point and the center point, wherein A represents the current center point, and P1-P8 represent adjacent points in 8 directions of the center point respectively;
performing horizontal traversal on the binary image once, and if the gray values of P3 and P7 are the same as the gray value of A, keeping the gray value of the point A unchanged; otherwise, if the value is 0, the gray value of the point A is not changed, and if the value is 1, the gray value of the point A is set to be 0; a vertical traversal of the image is made until there is no change in the gray value of the pixel.
5. The three-level classification fingerprint identification method according to claim 4, wherein the singular points of the fingerprint are located and extracted by: the method for extracting the singular points comprises the following steps:
(1) taking an 8-neighborhood graph of the pixel point as an example, making a closed curve surrounding the pixel point, selecting a certain point on the curve as a starting point P0 of the traversal pixel point, selecting the next pixel point P1 in a clockwise direction, and so on until the next pixel point returns to P0 again;
(2) taking the block direction values of the sub-blocks respectively taking two adjacent pixel points as the centers to make a difference, and recording the difference as delta (k);
(3) if-pi/2 < | delta (k) | < pi/2, then the same is maintained; if | Δ (k) | < = -pi/2, Δ (k) = Δ (k) + pi; otherwise, Δ (k) = Δ (k) -pi; next, calculating the sum of the direction differences on the closed curve;
(4) if = pi, it means that the surrounding curve includes a core point; if = -pi, it means that the surrounding curve includes a triangular point; storing the information of the singular points, such as coordinates, directions and types into a feature point set; if the two calculation results are not met, otherwise, no singular point exists in the region.
6. The three-level classification fingerprint identification method according to claim 5, wherein the fingerprint database is divided into several sub-classes:
(one) one-level classification based on pattern type
The fingerprint pattern is determined by the number and relative position of singular points, and the specific pattern is judged according to the included angle between the connecting line between the two points and the reference coordinate axis of the fingerprint image, and the specific method comprises the following steps:
(1) connecting the central point and the triangular point;
(2) when the number of the core points is more than or equal to 2, the core points are spiral lines;
(3) when the number of core points is less than 1, the shape is a bow shape;
(4) when the number of the core points is equal to 1 and the number of the triangle points is not equal to 1, the core points are in the miscellaneous class;
(5) when the number of the core points is equal to 1 and the number of the triangle points is also 1, calculating the direction difference theta between the line segment direction of the connecting line and the direction of the central point; if the Δ is more than pi/9, the shape is the right skip; if the Δ is less than-pi/9, the left skip is a left skip; if thetaj is less than or equal to pi/9, the ledger line belongs to;
two-stage classification of number of singularity-based inter-ridges
The second-stage classification mainly reduces the data quantity difference in each category, so the second-stage classification is carried out aiming at the left skip shape, the right skip shape and the spiral shape, the ridge line number is calculated based on the connecting line between singular points, and the number of the intersection points of the straight lines and the ridge lines is calculated by adopting the following method:
(1) traversing 8 neighborhoods of points on the connecting line along the connecting line direction by taking the triangular points or the core points as starting points, judging the number of ridge line pixel points, and if the number is less than 2, not calculating to be an intersection point; if the number is 2, executing the step (2); if the number is 3, executing the step (3);
(2) calculating the distance s between two ridge line pixel points, and if s =1, not calculating an intersection point; otherwise, the total number of the intersection points is added by 1;
(3) counting the number of the pixel points containing 3 ridge lines, and if the number is an even number, adding 1 to the total intersection point number;
(4) traversing the intersection point set, calculating the difference value of the horizontal and vertical coordinates of two adjacent intersection points, if one of the two difference values is equal to 1, selecting one of the two intersection points to be removed from the intersection point set, and subtracting one from the total number of the intersection points;
dividing the ridge number into 13 Ri types (1 < = i < = 13); setting the upper limit of the search radius as r, calculating the ridge number N between singular points of the fingerprint to be identified, and searching the homologous fingerprint in the database as follows:
(1) obtaining a corresponding category Ri according to the ridge line number N, and setting a search radius c = 0;
(2) searching fingerprints in the corresponding categories and sequentially comparing the fingerprints with the fingerprints to be identified, and finishing matching if the matching is successful; continuing the following steps if the unmatching is successful;
(3) gradually increasing the search radius by taking 1 as a gradient, repeating the steps (4) and (5) until the matching is finished when the search radius is larger than the upper limit of the search radius, and declaring the search to fail;
(4) when the sum of the category and the search radius is less than 14, the data are one by one at Ri+cComparing the fingerprint template in the category with the fingerprint to be identified, and if the matching is successful, ending the searching process; otherwise, executing the step (5);
(5) when the category is larger than the search radius, the category is one by one in Ri-cComparing the fingerprint template in the category with the fingerprint to be identified, and if the matching is successful, ending the search; otherwise, the matching fails;
(III) three-stage classification based on ridge line average period
The method comprises the steps of firstly constructing a 7 x 7 mean filter to perform convolution operation on image frequency values, correcting partial pixel points with the frequency value of 0 into pixel points with the frequency value, namely smoothing the image frequency to obtain a more accurate frequency map, after the frequency value is obtained, using the geometric center of a fingerprint effective area as a standard, sorting fingerprint images into 60 x80 images, calculating the frequency of each image and then dividing the frequency of each image by the size of a whole image 60 x80 to obtain a ridge average period value, firstly associating the average period with the number of fingerprints, and then dividing the ridge average period into 10 types.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111477702.5A CN114120378A (en) | 2021-12-06 | 2021-12-06 | Three-level classification fingerprint identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111477702.5A CN114120378A (en) | 2021-12-06 | 2021-12-06 | Three-level classification fingerprint identification method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114120378A true CN114120378A (en) | 2022-03-01 |
Family
ID=80366949
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111477702.5A Pending CN114120378A (en) | 2021-12-06 | 2021-12-06 | Three-level classification fingerprint identification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114120378A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115527244A (en) * | 2022-11-28 | 2022-12-27 | 深圳市航顺芯片技术研发有限公司 | Fingerprint image matching method and device, computer equipment and storage medium |
CN116665256A (en) * | 2023-06-05 | 2023-08-29 | 长春理工大学 | Fingerprint matching method based on fingerprint image local area quality |
CN117218691A (en) * | 2023-11-09 | 2023-12-12 | 四川酷比通信设备有限公司 | Unlocking method based on fingerprint identification |
CN117975513A (en) * | 2024-04-02 | 2024-05-03 | 艾信智能环境科技(无锡)有限公司 | Medical logistics site identity intelligent recognition system |
CN117993855A (en) * | 2024-02-02 | 2024-05-07 | 北京中友科技有限公司 | Power industry park personnel management software system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007107050A1 (en) * | 2006-03-23 | 2007-09-27 | Zksoftware Beijing Inc. | Fingerprint identification method and system |
CN101154263A (en) * | 2006-09-27 | 2008-04-02 | 中国科学院自动化研究所 | System and method for recognizing analog fingerprint of twins based on partial structure |
CN105975909A (en) * | 2016-04-27 | 2016-09-28 | 湖南工业大学 | Fingerprint classification method based on fractal dimension and fingerprint three-level classification method |
-
2021
- 2021-12-06 CN CN202111477702.5A patent/CN114120378A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007107050A1 (en) * | 2006-03-23 | 2007-09-27 | Zksoftware Beijing Inc. | Fingerprint identification method and system |
CN101154263A (en) * | 2006-09-27 | 2008-04-02 | 中国科学院自动化研究所 | System and method for recognizing analog fingerprint of twins based on partial structure |
CN105975909A (en) * | 2016-04-27 | 2016-09-28 | 湖南工业大学 | Fingerprint classification method based on fractal dimension and fingerprint three-level classification method |
Non-Patent Citations (2)
Title |
---|
郝培蓉: "基于分类特征的指纹识别***的研究与实现", 《中国优秀硕士学位论文电子期刊(全文数据库)信息科技辑》, 15 February 2021 (2021-02-15), pages 6 - 40 * |
陈英;: "自动指纹识别***的实现", 浙江科技学院学报, no. 04, 31 August 2012 (2012-08-31), pages 303 - 309 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115527244A (en) * | 2022-11-28 | 2022-12-27 | 深圳市航顺芯片技术研发有限公司 | Fingerprint image matching method and device, computer equipment and storage medium |
CN116665256A (en) * | 2023-06-05 | 2023-08-29 | 长春理工大学 | Fingerprint matching method based on fingerprint image local area quality |
CN116665256B (en) * | 2023-06-05 | 2024-03-15 | 长春理工大学 | Fingerprint matching method based on fingerprint image local area quality |
CN117218691A (en) * | 2023-11-09 | 2023-12-12 | 四川酷比通信设备有限公司 | Unlocking method based on fingerprint identification |
CN117218691B (en) * | 2023-11-09 | 2024-01-26 | 四川酷比通信设备有限公司 | Unlocking method based on fingerprint identification |
CN117993855A (en) * | 2024-02-02 | 2024-05-07 | 北京中友科技有限公司 | Power industry park personnel management software system |
CN117975513A (en) * | 2024-04-02 | 2024-05-03 | 艾信智能环境科技(无锡)有限公司 | Medical logistics site identity intelligent recognition system |
CN117975513B (en) * | 2024-04-02 | 2024-06-07 | 艾信智能环境科技(无锡)有限公司 | Medical logistics site identity intelligent recognition system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Jain et al. | Intelligent biometric techniques in fingerprint and face recognition | |
Halici et al. | Introduction to fingerprint recognition | |
CN114120378A (en) | Three-level classification fingerprint identification method | |
CN107748877B (en) | Fingerprint image identification method based on minutiae and textural features | |
CN100414558C (en) | Automatic fingerprint distinguishing system and method based on template learning | |
EP2174261B1 (en) | Fingerprint matching method and apparatus | |
Bansal et al. | Minutiae extraction from fingerprint images-a review | |
Jain et al. | Automated fingerprint identification and | |
Hambalık | Fingerprint recognition system using artificial neural network as feature extractor: design and performance evaluation | |
JP2827994B2 (en) | Fingerprint feature extraction device | |
Parkavi et al. | Multimodal biometrics for user authentication | |
US20080273769A1 (en) | Print matching method and system using direction images | |
CN111914755A (en) | Eight-direction gradient-solving fingerprint identification model | |
Li et al. | A fingerprint indexing scheme with robustness against sample translation and rotation | |
Nachar et al. | Hybrid minutiae and edge corners feature points for increased fingerprint recognition performance | |
Shin et al. | Detecting fingerprint minutiae by run length encoding scheme | |
Mali et al. | Fingerprint recognition using global and local structures | |
CN115795394A (en) | Biological feature fusion identity recognition method for hierarchical multi-modal and advanced incremental learning | |
Mane et al. | Novel multiple impression based multimodal fingerprint recognition system | |
CN115797987A (en) | Finger vein identification method based on joint loss and convolutional neural network | |
Dyre et al. | A survey on various approaches to fingerprint matching for personal verification and identification | |
Nirmalakumari et al. | Efficient minutiae matching algorithm for fingerprint recognition | |
Preetha et al. | Selection and extraction of optimized feature set from fingerprint biometrics-a review | |
CN109614915B (en) | Feature point positioning method | |
Palanichamy et al. | A novel image alignment and a fast efficient localized euclidean distance minutia matching algorithm for fingerprint recognition system. |
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 |