This is perhaps the oldest known identification technique that is still used on a large scale. Common examples are criminal records all over the world, forensic and access-control mechanisms in high-security establishments. Relatively simple to deploy and easier to use than other biometric techniques, fingerprint identification is based on the uniqueness of one’s fingerprint.
The principle is simple enough and involves a high-resolution snapshot of the fingerprint taken and stored in a database. This, however, seems to be a Herculean task. Imagine the number of unique fingerprints that might need to be stored and then accessed in a database. Mind-boggling! This necessitates the use of a classification mechanism to do a coarse classification of a database of prints for easier indexing and matching when needed. So, the whole set of prints is classified on the basis of whorls, arches and loops. This significantly reduces the time required to match a print as similar patterns go under a single class. This is done by means of algorithms that ‘extract’ these parameters from prints. The first step hence involves a coarse matching, followed by a fine matching of prints. Image enhancement techniques are also used during the process.
Two approaches are followed for getting the final match: minutiae-based recognition and correlation-based approach. While the first relies entirely on mapping the complete print, the second works on the gray-scale information of the print.
Look at an enlarged image of a fingerprint and you will find that it is made up of a series of ridges and furrows. These run parallel to each other and also form whorls, arches and loops. Minutiae points are characteristics of these ridges at endings and bifurcations. These points can be identified in a sample of a fingerprint. These are then matched against a database image. Depending upon a threshold number of minutiae points, the decision on a ‘match’ can be taken. The accuracy of this approach, however, depends on some factors. First is the quality of the fingerprints. Smudged prints and prints with a lot of ‘noise’ restrict the number of minutiae that can be uniquely identified. Also, identification points (minutiae) are not representative of a complete fingerprint.
Unlike the previous approach, the correlation method is more representative of the complete print. The identification system identifies ‘templates’ in the print. These templates are gray-scale data. A scheme of template matching is then used to identify matching positions in the print to be verified. Template data from these matching positions is then compared. The limitations of this scheme are basically because of orientation and location of the vertex for the template, which determines the directional gray-scale data. This will in turn determine that the identification hardware or software is ‘looking’ at the same orientation in the stored print and the print to be identified.
Developments in technology and methods of capturing fingerprints have improved the reliability of fingerprint-recognition systems. So we have small, pointable devices to capture prints and huge databases that can be searched within seconds to authenticate people. This will ensure that this field of biometrics remains popular.