Biometrics, being the scientific study of mathematical or statistical properties of human physiological and behavioral characteristics, is widely used in the sphere of information security (Jain, 1999). And the use of fingerprint as means of biometrics is one of the oldest methods of automatic identification and also the most common one in our time. The factors that promote the use of such systems are: small size and cost of equipment for processing of fingerprint images, high-performance hardware, power and speed of recognition, which meets the requirements of the software, a sharp growth and development of network technologies and the Internet, and also awareness of the need of simple, basic methods of protection and security of information (Genge, 2002).
This paper is therefore dedicated to the definition of the classical methods of identification of person by fingerprints and analysis of its various modifications. The major problems that arise at different stages in identification systems are also highlighted. This paper contains the statistics and the efficiency of recognition methods that help to determine the effectiveness of the methods for different systems (William, Herschel, Henry, 1974).
In our days, there are several methods for identification of a person by fingerprints. Each of the developed approaches has its own characteristics that are based on the properties and features of the fingerprint pattern. The basic idea of the work is to determine the optimal method of minimization of monetary and time costs and maximization of the degree of identity recognition. The main objective of the study is evaluating of the most common methods of identification by fingerprints.
Image processing and recognition
Upon the receipt of fingerprint images it is necessary to make its processing. During the processing, the image should achieve the best possible quality to reach the correct recognition. The main stages of image processing are: the elimination of interference (noise) and improvement of clarity, and identification of the main characteristics of the recognition. Because of the fact that the fingerprints may be contaminated, damaged, wet or dry, the problem of maximum filtering of image may arise in order to display images more clearly. For this purpose, two operations may be performed: matching and adaptive filtering and adaptive threshold segmentation. Despite the possible breakages and heterogeneity of individual fingerprint images, there can be determined their direction. Such filtering is applied to each image pixel. The results of all subsequent processing steps depend on the quality of the original image, which will be processed further. Therefore, any risk performed in order to increase the speed of image processing is usually unjustified. This will lead to a deterioration of recognition, which in turn will result in the need to repeat the verification system or recognition of incorrect data (Rudman, 1980).
It is known that the pictures of the skin covering vary. The palms contain speeches, called papillary lines and are separated by grooves, flexion lines, wrinkles and folds (white line), and pores. The most noticeable elements of the covering are flexion lines. White lines (wrinkles) appear due to loss of elasticity and skin dryness, and age-related changes. These lines usually play a supporting role during the identification process. The most significant papillary lines and pores have different shapes and are located at different distances from each other and from the edges of papillary lines. These lines on the palms and nail phalanges are quite complex and have diverse structure (Henry, 1990).
Basic properties of papillary pattern are the following: individuality, stability and recoverability. Individuality is that every man should tread pattern that is peculiar only to him. This is due to the peculiarities of anatomical structure and biological functions of the skin, as well as genetic peculiarity of man. Even identical twins’ set parts of skin pattern never repeats. For more than one hundred years of the practice there was no any single case of coincidence of all the details of each pattern in different people. Details are not repeated and even they are different at different fingers of the same person. According to mathematical calculations, the probability of coincidence of papillary patterns on all ten fingers on two people is so small, so it can be neglected (Rudman, 1980).
Stability means that the papillary lines appear on the 3-4th months of gestation of human and are stored until full putrefactive decomposition of the skin. With the growth of the organism, only dimensional characteristics change, but not the patterns. And finally, the recoverability guarantees full recovery of the pattern of damaged top layer of skin (epidermis). When a deep injury of the dermis happens or scars occur, they even increase the number of individualizing characteristics. An important characteristic of the skin is its ability to be displayed at those items that people touch. Moreover, it occurs independently of personal will and desire, but due to physiological properties of skin: its surface is always covered with secretions that stick to the surface (Henry, 1990).
Papillary patterns of nail phalanges are formed by three streams of papilla: the line center, periphery and base. Part of a pattern in which these flows unite forms a typical plot, called the delta, because it is similar to that Greek letters. Papillary patterns of nail phalanges are divided into types and species depending on the image to the center. For this reason, there are three types of patterns: arc, loop and curls (Genge, 2002).
Individual signs of papillary pattern are used for particular identification. They are some peculiarities in the structure of each papillary line of small morphological differences in detail. These include the eyes, islands, hooks, bridges, fragments, split (fork), the beginning of lines, scars, pores, breaks, bends, thickening, especially deltas, periods, merger of papillary lines and their fragments. For individual identification, comparable traces are unique sets of features matching in separate (Henry, 1990).
Traces of papillary patterns that are suitable for identification are static traces left on the smooth (polished) or plastic surface, the structure of which is much smaller than the features of papilla. Traces of fingerprints are surface and volume, visible and invisible, static and dynamic. Invisible traces remain on the objects’ surface which absorbs sweat and grease (paper, fabric, leather, cardboard, plywood, etc.). In the investigative and expert practice, often visual, physical and chemical methods of detecting traces of hands are used (Henry, 1990).
On stage of recognition, fingerprint that requires verification is compared to a reference. Typically, such recognition is based on comparing the similarity of neighboring features. Each of these features contains three or more adjacent characters, each of which is located at some distance, and has a definite orientation in relation to neighboring features. In addition, as it was mentioned, each feature is characterized by the unique type and direction of images, which are also compared. If after the recognition the very minor differences appear between neighboring features on the reference sample and that one presented for verification, then it is considered that they coincide. This is usually performed for all signs, and when it turns out that they coincide with a given accuracy, then we can talk about recognition of submitted patterns for fingerprint authentication (Henry, Gaensslen, 1994).
The result of recognition is to identify the number of identical signs. This is a number from 0 to n. The more matches exist; the more likelihood of recognition is present. It is a measure of the number of matches to set threshold. If this number is greater than the threshold, then there is a positive verification, that is, prints correspond to the reference, otherwise it is negative verification. Threshold value may also vary. There is a sense to increase the threshold for reliability of verification. Or, vice versa accordingly – for lower threshold decreases the reliability of verification and therefore decreases the number of mismatch with the standard (Boy Scouts of America, 1964).
Modifications of the classical approach
As one of the most difficult tasks of image processing of fingerprints is getting a clear image for recognition, there are several methods for its solution. Most of these methods use adaptive recognition of certain parts of the image. First, image is divided into squares that provide characteristics of the images and their orientation. Orientation of each section of the processing is determined in the process of spatial domain performance of two-dimensional fast Fourier transform (Henry, Gaensslen, 1994).
There is another approach, whereby the image of papilla is derived from the original input image in shades of gray. The result is an image with selected and split ends on the print as a result of the traditional processing of such depictions. Thus, instead of the use of one window (segment) to determine the location and orientation of the so-called papilla, there is a multi window mode. At first, the processing of the image of selected size is conducted, orientation of papilla is defined, their ends and splits are clearly distinguished. If this value is smaller than a certain threshold, then this window is divided into four smaller windows and the same is repeated for each windows. This method is used in order to avoid smoothing in some parts of the image that is often characteristic for the central part of the image (Genge, 2002).
Prints can be compared through correlation as well. To be precise, the correlation of two images includes transmitting of one image to another, so a couple of prints that match will have a higher correlative value. Threshold classification determines whether there is a sufficient amount of conformity in order to present fingerprint as recognized. Correlation analysis can be performed not only in spatial but also in frequency domain (Jain, 1999).
In addition to the classic approach of identification of the person through the fingerprint, there are methods that are based upon comparative analysis of methods for identification and verification of fingerprint images. The advantages and disadvantages of existing methods are highlighted, and problems that arise at different stages of image processing and process identification are identified.
It was concluded that it is difficult to select the best method because it is expedient to compare algorithms and hardware. For systems that require special security conditions multimodal biometrics should be used. The use of biometric authentication simplifies the identification of a person, and raises the reliability of safety systems.
Genge, N. E. (2002). The Science of Crime Scene Investigation. New York: The Ballantine Publishing Group.
Henry, E. R. (1990). Classification and Uses of Finger Prints. London: George Routledge and Sons Limited.
Henry C. Lee, R.E. Gaensslen. (1994). Advances in Fingerprint Technology. CRC Press.
Rudman, Jack. (1980). Fingerprint Technician. Natl Learning Corp.
Jain, Anil. (1999). Biometrics, Personal Identification in Networked Society : Personal Identification in Networked Society. Kluwer Academic Publishers.
Boy Scouts of America. (1964). Fingerprinting. New Brunswick, NJ.
William J., Sir Herschel, Edward R. Henry. (1974). The Origin of Finger-Printing Bound With Classification and Uses of Finger Prints. Ams Press, Inc.