Latent fingerprint identification system for crime scene investigation

Abstract

The traditional means of criminal investigation used in Nigeria is often unreliable and leads to innocent people's wrongful detention and a lack of justice for deserving offenders. The poor record-keeping and weaknesses in Nigeria's investigation process significantly contribute to the high levels of crime and insecurity in Nigeria. To tackle these issues, this project provides an implementation of a fingerprint identification system to improve criminal investigation in Nigeria. Three image processing algorithms and a Convolutional Neural Network classification algorithm were explored for matching performance. The Convolutional Neural Network classification model performed better than the three image processing algorithms with an accuracy of 64.44%. The final system provides a web interface with database interaction to send a fingerprint image and meta data to receive match results and potential suspect (criminal) information.