RU2305317C2 - Method of comparison of papillary patterns - Google Patents

Abstract

FIELD: identification of prints of papillary patterns.

SUBSTANCE: method can be used in forensics, in unauthorized access preventing systems, for card user identification on base of papillary patterns of feet, fingers and palms. Nest is built for any peculiarity, which nest has geometrical and morphological characteristics of vicinity. Then nests with different patterns are subject to comparison and best pairs of nests are selected. Fragments of patterns to be compared are developed from any selected pair of nests by means of transition from one pair of nests to other one along way of best comparison of nests. Then estimations of comparison are stored for any way of development of fragments and the vest way is selected. Estimations of comparison of nests are subject to correction depending on degree of similarity of areas of patterns which patterns do not contain peculiarities.

EFFECT: reduced chance of error of identification.

2 cl, 5 dwg

Description

The invention relates to the field of forensics and is intended to compare fingerprints, palms and feet. Its use allows to obtain a technical result in the form of reducing the likelihood of identification errors of the papillary pattern.

A known system for recognizing and searching for fingerprints, in which the papillary pattern is scanned by a rotating line around the center of the pattern and when the line meets a feature, a mathematical code is recorded containing the feature type, ridge count, angular coordinate and distance (PCT application No. 87/01224, IPC G06K 9/00, publ. 1987).

The disadvantage of the system is the dependence of the mathematical code on the traversal order of features, which is violated when the coordinate system is shifted, as well as the high error in measuring the crest count at a small angle of intersection of the scanning line with the papillary line.

A known method of comparing fingerprints based on a network model of features, the nodes of which are described by coordinates and direction. When comparing two network models (query and archive fingerprint), a list of similar pairs of model nodes is formed, then coordinate systems are optimized and, comparing the nodes of the two models, the degree of similarity of the models is determined (US patent No. 4646352, IPC G06K 9/68, publ. 1983).

This method offers a network model, which is highly dependent on wandering edge features, which increases the uncertainty of the network model and increases the likelihood of identification errors of papillary patterns.

A known method of comparing papillary finger patterns, in which features are recorded in the area of a clear fingerprint with their coordinates, direction, type and comb counting to other features, features of different fingerprints compare and form a compatibility matrix of pairs of features, from which the best unique combination of identical features is extracted request and archive fingerprints (RF application No. 94039018, IPC G06K 9/00, 9/68, publ. 1996, bull. No. 25).

This method is based on an estimate of the compactness of the set of points on the differential plane. However, with elastic deformation of one of the compared fingerprints, the compactness of the set of points on the differential plane is violated. Moreover, the estimation of compactness depends on the choice of the center of coordinates, which leads to uncertainty and reduces the reliability and selectivity of the identification of papillary patterns.

Closest to the invention is a method for comparing papillary finger patterns, which consists in registering numbered features in a clear imprint area, determining for each feature coordinates, type and direction vector, ridge count between features, a nest consisting of ordered pairs that are represented by a number-event and by the number of the features associated with the event, assessing the degree of similarity of nests from different patterns, compiling a list of pairs of similar nests according to the results of the best estimates, parallel to development of fragments from different patterns from each pair of nests from the list by the method of transition from one pair of nests to another along the path of the best comparison of nests, evaluating each development path of fragments and choosing the best rating (Russian patent No. 2185661, IPC7 G06K 9/62, publ. 2002) . This method is selected as a prototype.

The disadvantage of the prototype is that the method does not analyze areas of the papillary pattern with clear and distinct papillary lines, but without features (hereinafter free areas). These areas are the same important identification feature as the features themselves, and the use of free areas of the pattern allows you to filter out candidate patterns that clearly contradict the pattern in terms of size and shape of free areas.

The problem solved by the present invention is to reduce the likelihood of identification errors of the papillary pattern.

This is achieved by the fact that in the region of a clear imprint of a papillary pattern, papillary lines and features are distinguished, encoded features, for each feature a nest is formed in the form of ordered pairs of events and feature numbers, on the compared papillary patterns in the area of a clear imprint, areas containing no features are outlined, evaluated the parameters of the outlined areas, in parallel, develop fragments from different papillary patterns by the method of transition from one compared pair of nests to another along the path of the best assessment of similarity to anger d accumulate similarity evaluation jacks, jacks refine estimates of similarity in the development patterns of fragments in accordance with the degree of similarity parameters outlined areas and their location on the compared patterns.

In addition, the nests of the compared prints are compared after they are geometrically tied to loops, deltas, and curls.

The essence of the proposed method is illustrated by five drawings, which show:

figure 1 between two papillary lines shows a feature of the type of ending with a direction vector;

figure 2 between two papillary lines shows a feature of the type of branching with a direction vector;

figure 3 for features 1 of the type of ending shows a socket, which is formed by features 2-5;

figure 4 shows the paths of parallel development of fragments along nests from different papillary patterns;

figure 5 shows a fingerprint with an informative and non-informative area, in the informative area there are features with vectors of directions and circles of different radii covering that informative region of the pattern that does not contain a feature.

Consider the sequence of actions in the present method.

Papillary lines stand out on the fingerprint, palm or foot. The areas of the imprint with clear papillary lines are called the informative area, and the greases, imperfections, burns, the trace-forming surface are called the non-informative area. On clear papillary lines, features are detected. Free areas are formed in the informative area of the fingerprint, which does not contain features. Each feature in the field of a clear papillary pattern, as in the prototype, is numbered and described by coordinates, a direction vector in the direction of increasing the number of papillary lines, type of feature, comb counting to other features, scale characteristic. For each feature, a nest is formed consisting of ordered pairs, each pair being represented as an event and the number of the feature associated with the event. The event in each pair is described by a number depending either on the type of feature located on the line, or on whether the feature is located to the right or left of the line.

Consider the features of two types: branching and ending (Fig.1, 2). For each feature, you can synthesize a nest. For example, a nest on a papillary pattern for a feature 1 of the ending type (Fig. 3), which includes several nearest other features 2-5, selected, say, priority along the lines, is described by a matrix of the form

where d _ij is the distance from the i-th feature to the j-th feature, α _ij is the minimum rotation of the direction vector of the i-th feature to the direction vector of the j-th feature, ρ _ij is the ridge count between the i-th feature and the j-th feature , e _ij is the event in the slot of the ith feature for the jth feature, n _j is the number of the jth feature. The composition of the nest can be expanded, although this is not essential for the description of the method. The set of parameters d _ij , α _ij , ρ _ij , e _ij , n _j is called the connection between the i-th and j-th features. Note that the communication parameters are independent of the orientation and choice of the center of coordinates.

Such nests are built for each feature.

Compare the nest Г _{k of} one pattern with the nest Г _{l of} another pattern (Fig. 4). To do this, we choose the first connection of the nest of G _k and compare it with all the connections of the nest of G _l by the correlation method. We choose the most similar connection of the socket Г _l , remember the result of the comparison with the connection of the socket Г _k , and delete the resulting pair of links from further consideration. We perform similar operations sequentially for all the connections of the jack Г _k . The averaged correlation value of isolated pairs of bonds is an estimate of the comparison of two nests. The method for comparing two nests can be improved using the Bayesian approach or the Neumann-Pearson method.

Adjust the assessment of nests G _k and G _l . To do this, outline the free areas of the compared patterns (figure 4). A method for describing free areas is presented in Fig. 5, where the bold rectangle limits the region of the imprint, inside which the informative region of the pattern is framed by a dashed curve. Inside the informative area of the pattern, there are features indicating the direction vectors along which the left loop is guessed. Some features are far enough from each other, and so much so that circles of different radii can be entered in the informative area (without going beyond it) so that they do not cover the features. Usually, large circles are entered at first, and then into the free areas uncovered by circles, smaller ones. The circles in FIG. 5 are indicated by a dotted line. Many circles with a small mutual intersection cover free areas. Although the maximum and minimum radii of the circles are selected during the setup of the algorithm, the minimum radius should not be less than the width of 2-4 papillary lines. As a result, we get many circles with a known location and radius. The method for describing free areas is not unique.

Each free region is described by parameters such as area, center of gravity, degree of elongation, amount of bending ... which are calculated from the analysis of intersecting circles (known methods are described in mathematics manuals, for example, calculation of the center of gravity). 4, the centers of gravity of the free areas are shown in bold circles. If for the slot Г _k there is a nearby free region with the parameters {p _i }, then for the slot Г _{l of} another print something similar should be expected (Fig. 4). In the absence of a free region adjacent to the slot Г _{l, the} comparison estimate of two nests is underestimated by multiplying the estimate by a coefficient less than 1. If there is a similar free region, the comparison estimate of two nests is multiplied by the correlation coefficient of the parameters {p _i } of the corresponding free regions from different patterns. With the complete similarity of the free regions, the estimation of the comparison of the two nests does not change (the correlation coefficient is 1).

Since each nest of one pattern is compared with the nests of another pattern, according to the selected best estimates of the similarity of the nests, several candidates are selected — nests from another pattern, which may differ from a similar set of candidates obtained by the prototype method. Received pairs of nests form a list of pairs of nests. Of all the received pairs of nests, some of the best pairs of nests are selected. This allows you to speed up the pattern comparison process, although shortening the list of pairs of nests is optional and sometimes risky.

Perform the development of fragments on the compared patterns. To do this, select one pair of nests from the list of pairs of nests: nest Г _{k of} one pattern and the candidate corresponding to this nest - nest Г _{l of} another pattern. We choose the first connection of the nests of G _k and remember the correlation with the first connection of the nests of _l . At the ends of these connections are other nests, which we call incidental to these connections. The assessment of the incident pair of nests is already determined in the previous step. We multiply the estimate of the incident pair of nests, corrected by the degree of similarity of the free areas of the patterns being compared, by the estimate of the corresponding pair of bonds and obtain an estimate of the step of the path. We perform similar operations for all the connections of the socket Г _k , and all the connections of the socket Г _l by brute force and choose the best step of the path. Remember the assessment of the best step of the path. Perform the transition corresponding to the best step of the path to the incident nests: from the nest Г _k to the nest Г _{k + 1 of} one fingerprint and from the nest Г _l to the nest Г _{l + 1 of} another fingerprint (Fig. 4), and the pair of nests Г _k and Г _l Cross out from further consideration.

Consecutively making transitions from one pair of nests to another, we construct a path whose number of vertices does not exceed the minimum number of features of one of the prints. The development of this path is the synchronous development of fragments starting from the starting nests Г _k and Г _l on the patterns being compared. An estimate of the entire path can be obtained as an average estimate of the best steps of the path.

We perform the same operations for the development of other fragments starting with other pairs of nests from the list of pairs of nests. From the estimates of all the paths for the development of fragments, we choose the best estimate, which we will take as the result of comparing the two patterns.

This completes the comparison of papillary patterns.

The method for developing fragments can be improved using well-known optimization methods. It is clear that the exhaustive procedure for constructing a list of pairs of nests is slow. It can be accelerated with loops, deltas and curls. Since the loops and deltas are oriented and their location is known, it is possible to calculate geometric sectors corresponding to approximately the same areas of the patterns being compared, in which the same features should be located. Thus, for each nest of one pattern, the possible number of candidate nests from another pattern is reduced.

Note the advantages of the proposed method.

The present invention is not inferior in accuracy of identification of papillary patterns to the prototype, since it almost repeats the prototype algorithm. However, taking into account free areas allows us to refine the estimates of similarity of nests, additionally sifting out “foreign” nests or recognizing “native” ones, and thereby reduce the likelihood of identification errors of papillary patterns by correcting the estimates of the best path.

This method of comparing papillary patterns is implemented in the form of a sequence of actions performed on a computer and is used in forensics, in systems for preventing unauthorized access to secure premises, when identifying the identity of a person using a credit card.

Claims (2)

1. A method of comparing papillary patterns, which consists in registering numbered features in the clear imprint of the papillary pattern, determining for each feature coordinates, direction vector, type, comb count to other features, forming a nest containing at least two features and consisting of ordered pairs, which are represented by the event-number and number of the feature associated with the event, comparing the nests of one pattern with the nests of another pattern, building a list of pairs of similar nests, parallel the development of fragments on both patterns from each pair of similar nests from the list by the method of moving from one pair of similar nests to another along the path of the most similar nest, choosing an assessment of the development of fragments of patterns, characterized in that on the compared papillary patterns in the region of a clear imprint outlines areas that do not containing features, evaluate the parameters of the outlined areas such as area, center of gravity, elongation, the amount of curvature, and refine the results of comparison of nests during the development of fragments of patterns in accordance and with the degree of similarity of the parameters of the outlined regions and their location on the patterns to be compared, in the presence of a similar free region, the comparison estimate of two nests is multiplied by the correlation coefficient of the parameters of the corresponding free regions from different patterns, with the complete similarity of the free regions, the comparison comparison of the two nests is not changed and the correlation coefficient equal to unity, and in the absence of a free region adjacent to the nest, the indicated assessment of the comparison of two nests is underestimated taking into account the correlation coefficient. 2. The method of comparing papillary patterns according to claim 1, characterized in that the nests of the compared prints are compared after they are geometrically linked to loops, deltas and curls.

Images