RU2821145C1 - Method for biometric personal identification by hand parameters - Google Patents

Abstract

FIELD: physics; pattern recognition.

SUBSTANCE: invention relates to pattern recognition. Disclosed is a method in which a reference image is recorded in a given scale and a given angle of the hand silhouette of a certain person with fingers apart, recording in the same scale and the same angle an image of the hand silhouette of the hand of the person to be identified with fingers apart, and after processing said registered images, their parameters are compared. Task is solved due to the fact that the shapes of the contours of the fingers are compared without any transformations. For quantitative estimation of contour difference the square of Euclidean distance is used.

EFFECT: invention makes it possible to reduce probability of incorrect decision making by identification device, id est reduce probability of missed intruder and false alarm.

10 cl, 4 dwg

Description

Technical field

The invention relates to image recognition, personal identification, individual registration of entry and exit, control of access to the territory, buildings, premises and objects, devices, computer programs, etc.

More specifically, the invention can be used in biometric identification devices to control access to objects based on hand parameters.

Prior Art

The quality of operation of biometric identification devices is characterized by the likelihood of making erroneous decisions: missing an intruder and a false alarm (false denial of access). There are known methods for biometric identification of a person based on the parameters of the hand, which consists in first measuring in a fixed position of the fingers and hand some sizes of specified parts of the hand of a certain person, then measuring in the same fixed position the same sizes of the same parts of the hand of an identified person and compare the indicated sizes.

This method is implemented in a device selected as an analogue of patent US 4736203 A, application 756390, priority 07/17/1985, published 04/05/1988; IPC: H04Q 1/00, Author: David P. Sidlauskas, Patentee: Recognition Systems, Inc., San Jose, Calif.

An identification device containing: a means that responds to retroreflection; a pair of orthogonally oriented retroreflective surfaces in front of which a human hand can be placed to retroreflect incident radiation from a radiation source without being interrupted by said hand to project a single image comprising a top view and a profile of said human hand to said retroreflective response means; means in said retro-reflective means that respond to said image to provide signals corresponding to said plan and profile views; means for responding to said signals for providing current measurement data from said plan and profile views; means for comparing said current measurement data with data previously stored in the memory; and means for providing a signal when said current measurement data and said previously stored data match within a predetermined tolerance.

The essential features of the analogue, which coincide with the features of the claimed invention, are as follows.

An image of a hand is recorded, the silhouette of the hand is isolated, compared parameters are obtained, and compared with similar parameters previously stored in memory.

The disadvantages of the analogue include the need to rigidly fix the position of the fingers and hand on the support, which causes certain inconvenience, and insufficient reliability of recognition due to measuring only the sizes of parts of the hand (such as length and width) and calculating the area, perimeter and the relationship between them.

There is a known analogue, described in the article “Using B-Spline Curves for Hand Recognition”, authors: Ying Liang Ma, Frank Pollick {ma, frank}@psy.gla.ac. uk, W. Terry Hewitt w.t.hewitt@man.ac.uk; Psychology Department, University of Glasgow, G12 8QB, UK, Manchester Visualization Centre, University of Manchester, M13 9PL, UK, published in 0-7695-2128-2/04 $20.00 (C) 2004 IEEE.

A method for biometric identification of a person based on hand parameters, which consists in using 4 B-spline curves to approximate the contours of the fingers (except the thumb) from one hand image for one person. These 4 curves, as well as other geometric measurements of the hand, are then stored as that person's "signature" in a database. By calculating the differences between the curves of the hand images from the database and the curves from the acquired hand image using the point projection method, the person is verified/identified by identifying the hand image in the database that is closest to the image of the presented hand.

The essential features of the analogue, which coincide with the features of the claimed invention, are as follows: images of the hand of an identified person with fingers spread are recorded, the contours of the fingers of the identified person are identified and compared with the contours of the fingers of a certain person previously stored in memory.

The disadvantages of this method include the following. Contour approximation is used, which is performed with errors and requires complex calculations. A single image is used to obtain the template. Errors are 3...5% based on a user base of 20 people. For comparison: GOST 52860 “Technical means of physical protection. General Technical Requirements" sets the maximum permissible probability of a false alarm and missing an intruder at 0.5% for the internal zone, and for a particularly important zone, the probability of missing an intruder should not exceed 0.1%.

The “Method for biometric identification of a person based on hand parameters”, described in RF patent No. 2304806, G06K 9/00, (2006.01), Application 2006102716/09, 01/31/2006, publ. 08/20/2007; author: Mestetskiy Leonid Moiseevich, patent holder: Limited Liability Company "Biometric Systems and Technologies" (RU).

A method for biometric identification of a person based on the parameters of the hand, which consists in first registering at least one reference image in a given scale and a given angle of the silhouette of the hand of a certain person with fingers apart, then registering an image of the silhouette of the hand in the same scale and the same angle hands of an identified person with fingers apart and compare the parameters of the specified registered images after their processing, characterized in that a number of circles are inscribed in each specified registered image, starting from the fingertips of the hand, lines are drawn through the centers of the specified inscribed circles, separately from the specified registered images of silhouettes the hands are straightened by the indicated lines drawn through the centers of the indicated circles inscribed in the images of the silhouettes of the corresponding fingers, through the indicated centers lying on the indicated straightened lines, straight segments are drawn perpendicular to the latter, equal in length to the radii of the corresponding indicated circles, through the ends of the indicated straight segments are drawn the compared lines combine the beginnings of the specified straightened lines of the corresponding fingers and determine the mismatch values of the compared lines of at least some of the corresponding fingers.

The essential features of the prototype, which coincide with the features of the claimed invention, are as follows: first, at least one reference image is registered at a given scale and a given angle of the silhouette of the hand of a certain person with fingers apart, then an image of the silhouette of the hand of an identified person is registered at the same scale and the same angle individuals with spread fingers and compare the parameters of the specified registered images after their processing.

The disadvantages of this method include the following. It is not the real contours of the fingers that are compared, but the figures obtained from the real contours, which coincide with the real contours only in width in a certain number of sections; therefore, the method does not allow distinguishing fingers that have the same width in the corresponding sections, but differ in curvature. When inscribing circles, local convex areas of the contour are smoothed out, therefore, some of the individual features of the contour are lost. These shortcomings increase the likelihood of missing an intruder. Using only one image to obtain a standard increases the likelihood of a false alarm, since a standard may be obtained that has parameters with the greatest differences from the parameters that will be implemented most often and can be considered a mathematical expectation.

Disclosure of the invention

The problem to which the claimed invention is aimed is to reduce the likelihood of the identification device making an incorrect decision.

The technical result of the claimed invention consists in comparing the real contours of the fingers without transforming the shape of the contour.

The technical result is achieved in that in the method of biometric identification of a person based on the parameters of the hand, which consists in registering at least one reference image of a hand with fingers apart, belonging to a certain person, obtained in a given scale and a given angle, registration belonging to an identified person in the same scale and the same angle of the image of the hand with spread fingers, obtaining compared parameters from the specified registered images, comparing the parameters of the specified registered images, according to the invention, from the above images, the contours of the hand are isolated, on the contours for each side of each finger, an area limited by the points located near the ends of the fingers and interdigital spaces. At each site, a set of points located at equal distances from each other is selected. Data about sets of points are generated in a form that uniquely determines the relative position of points in the set. Selected sets of points are approximated by straight lines. The angles between straight lines approximating sets of points on the two sides of the finger are determined. On each approximating straight line, select the point closest to the point limiting the contour section from the end of the finger, and determine the distance between the selected points for the two sides of each finger. The difference between the corresponding sets of points of a certain person and an identifiable person is determined. The difference in angles between straight lines approximating sets of points on the two sides of the corresponding fingers is determined for a specific person and an identifiable person. The difference in distances between the points on the approximating straight lines closest to the points limiting the contour sections from the ends of the fingers is determined for a certain person and an identifiable person.

Brief description of drawings

In fig. Figure 1 shows an image of a hand.

In fig. 2 shows the outline of the hand for the image in FIG. 1.

In fig. Figure 3 shows areas of finger contours used for identification.

In fig. Figure 4 shows sets of points for one of the fingers used for identification, as well as additional biometric parameters.

Carrying out the invention

The method of biometric identification of a person based on the parameters of the hand consists in the fact that first, at least one image of the hand with spread fingers, belonging to a certain person, obtained in a given scale and a given angle (Fig. 1) is registered. The image of the thumb is not used because the shape of the outline of the thumb can vary greatly. Then, an image of a hand with spread fingers belonging to the identified person is recorded at the same scale and from the same angle. For each image, the parameters to be compared are obtained and the parameters of the specified registered images are compared.

The compared parameters are obtained as follows. From each image, by any known method, for example, threshold or gradient, the contour 1 of the hand is isolated (Fig. 2) in the form of points belonging to the contour and located at a distance convenient for use in the identification algorithm (for example, 1 mm).

On contour 1 (Fig. 2), for each side of each finger, a section 2 is allocated (Fig. 3), limited by points 3 located at the ends of the fingers, and points 4 located at the ends of the interdigital spaces for the middle, ring and inner sides of the little and index fingers finger On the outer sides of the index finger and little finger, points 4 are selected so that the length of sections 2 on them is equal to the length of section 2 on the contours of the inner sides of these fingers or slightly longer. The points on the contour areas are selected so that the area is as large as possible, but does not include the area (for example, 8 mm from the end of the finger) where the nail plate can protrude beyond the border of the finger, and the area (for example, 5 mm) adjacent to the interdigital space , since the contours in these zones have an unstable shape.

At each site, a set of points is selected (Fig. 4), located at an equal distance d from each other. Specify either the distances between adjacent points or the number of points in sets. Then they generate data about the sets of these points in a form that uniquely determines the relative position of the points in each set, for example, they indicate the distances Ln of the points in the set from straight line 5, which approximates the set of points using the least squares method and given by the formula

Parameters a and b are calculated using known formulas:

where x _i and y _i are the coordinates of the i-th point in the set, n is the number of points in the set.

To quantify the differences in the shapes of sections of the contours of the fingers of a certain person and an identified person, given by sets of points with a fixed distance between neighboring points, the square of the Euclidean distance in the n-dimensional space between the sets is determined, where n is the number of points in the set.

Before calculating the distances, the sets of points for each side of each finger of the identified person and the identified person are positioned relative to each other in such a way that the square of the Euclidean distance is minimal. To do this, first place the points of the compared sets in the same coordinate system so that the ordinates of the points closest to the end of the finger coincide. For example, the distances of points from approximating lines are used as abscissas. Then new coordinates of the points of one of the sets (for example, an identified person) are calculated, at which the relative position of the points in the set will be preserved, and the square of the Euclidean distance between the sets will be minimal. To do this, it is necessary to rotate the set of points of the identified person by a certain angle and shift by a certain amount along the abscissa axis. To simplify the calculations, it is assumed that the ordinates of the corresponding points in the corresponding sets of a certain person and the identified person coincide and at small angles of rotation the ordinates y _i of the points of the set of the identified person do not change. New values calculated by the formula

The end values are found from the minimum sum condition

according to the formulas:

Where - coordinates of the i-th point of the reference set.

Then this sum itself is found (the square of the Euclidean distance, ) according to formula (10).

where x _i are the abscissas of the points of the reference set,

- abscissa of dialing points of an identifiable person.

If it is necessary to simplify calculations, the following value is used as a measure of the difference between sets of points:

Where p _i and q _i are the distances of the corresponding points of the corresponding sets of points of the identified person and a certain person from the straight line approximating the set using the least squares method.

The sets of points for the finger sides of an identified person can be considered as n-dimensional random vectors, and the reference sets obtained from the image of the hand of a certain person can be considered as mathematical expectations. To estimate the difference between a set of points on the finger side of an identified person and a reference set, squared distances between sets of points can be used, correlated with the number of points in the set (E.S. Ventzel, Theory of Probability, fourth edition, stereotypical, NAUKA publishing house, M., 1969 , pp. 315, 316):

statistical variance

or

• corrected (unbiased) variance

The resulting value is compared with a given threshold value.

In addition, the difference in angles α between the straight lines 5, approximating the sets of points on the two sides of the corresponding fingers for a certain person and the identified person, is determined, as well as the difference in the distances W between the points 6 on the approximating straight lines 5, closest to the points 3, limiting the sections of the contours from the ends of the fingers , for a specific person and an identifiable person.

Additional biometric parameters can be the lengths of selected sections of contours on the sides of the fingers. You can proceed as follows: when determining the differences between the sets of points obtained from the image of the hand of a certain person and the corresponding sets of points obtained from the image of the hand of the identified person, compare the lengths of selected sections of the contours on the sides of the fingers and if the length of any section of the contour of the identified person differs from the length of the corresponding section of the contour of a certain person by more than a given amount, then they conclude that the hand does not belong to a certain person and do not make other comparisons.

Scores can be used separately for each side of each finger and deny access if a specified threshold is exceeded for at least one side of one finger. The sum of deviations for the two sides of the finger can be calculated and compared with a threshold value. The sum of deviations for all fingers can be calculated and compared with the threshold value. The deviation values for the angles between the approximating lines and for the distances between points on these lines can be used separately, or added with weighting coefficients to the distances between sets of points.

None of the methods for selecting contour points that limit contour areas on the sides of the fingers guarantees that each time the hand of the same person is presented, the same limiting points will be selected, therefore additional comparisons of contour areas obtained from the image of the hand of the identified person are carried out individuals displaced along the contour relative to the initially selected areas by a given amount (for example, 1 mm and 2 mm) in one direction or the other. For identification, the smallest difference values are used.

The proposed method compares real finger contour shapes without any transformations. The method makes it possible to distinguish fingers that have the same width at given points, but differ in shape, and therefore reduces the likelihood of missing an intruder.

To obtain reference sets of points, the following method is also possible. Several images of a certain person are registered, sets of points on the contour are selected from each image, compared with the corresponding sets of points of other images, and sets of points that have the least amount of differences from the other sets are selected as reference ones. This method of obtaining reference sets of points reduces the likelihood of false alarms.

Thus, the likelihood of the identification device making an incorrect decision is reduced, i.e. the task is being solved.

The presence in the claimed invention of features that distinguish it from the prototype allows us to consider it to meet the “Novelty” condition.

New features were not identified in technical solutions for similar purposes. On this basis, we can conclude that the claimed invention complies with the “inventive step” condition.

Industrial applicability

The RFNC-VNIITF access control system uses an identification device based on the geometry of the hand called “GEOR”, which uses the method described above. The use confirmed the technical result stated here.

The information presented indicates that the claimed invention meets the “Industrial applicability” condition.

Claims (10)

1. A method for biometric identification of a person based on the parameters of the hand, which consists in registering at least one image of a hand with spread fingers belonging to a certain person, and registering an image of a hand with spread fingers belonging to the identified person in the same scale and from the same angle, for each image, the compared parameters are determined, the parameters of the specified registered images are compared, characterized in that the contours of the hand are isolated from the above images; on the contours of the hand, for each side of each finger, an area is selected, limited by points located at the ends of the fingers and interdigital spaces, and the compared the parameters include sets of points selected at each site, located at equal distances from each other, form data about sets of points in a form that uniquely determines the relative position of points in the sets, determine the difference between sets of points obtained from the image of a certain person’s hand, and corresponding sets of points obtained from the image of the hand of the identified person. 2. The method according to claim 1, characterized in that when determining the differences between sets of points, they are positioned relative to each other in such a way that the squares of the Euclidean distances between the corresponding sets are minimal, and they are determined. 3. The method according to claim 1, characterized in that when determining the differences between sets of points, the distances of each point from the straight line approximating the set of points are determined using the least squares method, and for each pair of sets of points of a certain person and an identifiable person, the sum of squared differences of the above distances is determined corresponding points of a certain person and an identifiable person. 4. The method according to claims 2 and 3, characterized in that when determining the differences between sets of points, the values characterizing the differences between sets of points are divided by the number of points in the set. 5. The method according to claims 2 and 3, characterized in that when determining the differences between sets of points, the values characterizing the differences between sets of points are divided by the number of points in the set minus one. 6. The method according to claim 1, characterized in that when determining the differences between sets of points, the difference in angles is determined between straight lines approximating the sets of points on the two sides of the corresponding fingers for a certain person and an identifiable person, 7. The method according to claim 1, characterized in that when determining the differences between sets of points, the difference in distances is determined between the points on the approximating straight lines of the two sides of each finger, closest to the points limiting the sections of the contour from the end of the finger. 8. The method according to claim 1, characterized in that when determining the differences between sets of points, the difference in the number of points in the set is taken into account if the distance between the points is specified. 9. The method according to claim 1, characterized in that when generating data on sets of points from each image, sets of points on the contour are selected, compared with the corresponding sets of points of other images and those sets that have the smallest amount of differences are selected as reference sets of points from other sets. 10. The method according to claim 1, characterized in that when determining differences between sets of points, the lengths of selected sections of contours on the sides of the fingers are compared, and if the length of any section of the contour of an identified person differs from the length of the corresponding section of the contour of a certain person by more than a threshold value , then they conclude that the hand does not belong to a certain person and do not make other comparisons.