Separation of Overlapped Fingerprint Images using Deep Learning

Criminal investigations have evolved with advancements in technology. Traces of fingerprints play a vital role in identifying and understanding crime scenes. These traces can form a clue to various suspects of crime. Fingerprints obtained from such crime scenes are often low quality, damaged and, most of the time, overlapped. It is a crucial requirement to separate overlapped fingerprints, which can be beneficial in identifying an individual, especially in criminal investigations, as it can approve or disapprove a person’s identity. Overlapped fingerprint separation is a challenging task. Several methods have been utilized to segregate overlapped fingerprints, resulting in the successful ability of the Automated Fingerprint Identification System (AFIS) to match the individual fingerprints.The objective of the proposed study is to classify the fingerprint images that are overlapped and separate these images by using a deep learning approach. Further, the result is extended to reconstruct these separated fingerprint images. The experimental results show that overlapped fingerprint images are efficiently separated, but there is sufficient scope for improving the reconstructed images of the damaged part of the fingerprint image.