Fingerprint Image Recognition for Crime Detection

Abstract

Fingerprint images play an important role in solving serial crime cases. The main objective of this paper is to provide a complete fingerprint identification analysis of crime scenes using deep machine learning which a class of Convolutional Neural Network is. Images taken as a database are generally insufficient and are difficult to classify. Therefore we use appropriate enhancement techniques for pre-processing the fingerprints which we have taken from the database. Minutiae are uprooted from the fingerprint images with the process of minutiae extraction. The preprocessed data is given as input to the CNN for training and testing. It further goes with the similarity checks. The experimental results demonstrated on the database using Matlab show high accuracy. Fingerprint images are important tools for finding culprits at the scene of the crime. Thus, the identification of the suspect using the fingerprints will be more quick and accurate. And we came up with an algorithm of deep learning architecture called Alexnet. This technology shows that it is feasible to get accurate results to identify the artifacts that we generally use in the actual scenario. This procedure offers up to 99 percent accuracy with much less than 6 seconds of classification time.