Evaluation of Existing Methods for Earprint Recognition

Abstract

Ear as a part of human body has been used in forensic practice but the use of earprints as evidence in criminal trials remains arguable. The Forensic Ear Identification (FearID) research project was started in order to study the strength of evidence of earprints found on crime scenes. A limited number of publications exist related to computerized methods used for earprint identification. The study presented here compares existing methods for earprint image identification. Two of them are point based methods that use Scale Invariant Feature Transform (SIFT) and Curvature Scale Space (CSS) feature for recognition process. The other is based on the image intensity value that compares two images by registration algorithm. We applied two different similarity metrics for evaluating registration. All methods carried out on a subset of FearID database. Equal error rate and hitlist behavior on our small dataset show that CSS is not useful in earprint recognition whereas SIFT and image registration technique have promising results.