Automatic fingerprint matching

Abstract

In automatic fingerprint identification systems poor quality impressions and latents often require expert interpretation if usable encodings are to be obtained. When this is necessary there are basically two different approaches taken for data entry "direct" encoding of the minutiae and "tracing" the ridges. The minutiae identified using these two methods on a test set of 49 latents were compared. The results indicate that these two methods are not just alternative ways of entering information but are radically different approaches which often lead to substantially different interpretations of the impression. The results of running searches on the two sets of minutiae using the Home Office matching algorithm do not indicate that either method is to be preferred from the point of view of accuracy. This conclusion is reinforced when the encodings are compared directly by eye with the matching impressions. Interpreting and encoding an impression viewed in isolation, by whatever method, is a fundamentally different task to the iterative process used by fingerprint experts to determining a match. There is scope for improving the matching accuracy of automatic fingerprint systems. The view is presented that a significant improvement could be achieved if algorithms are developed which are capable of iterative interpretation of the encoded or original data.