Fingerprint Image Enhancement for Crime Detection Using Deep Learning

Fingerprint has been most popularly used in many commercial applications for person identification. Latent fingerprints are produced largely via the finger sweat or oil left outs by the suspects accidentally. These impressions are generally blurred plus not observed by naked eye. These fingerprint images in crime prospects are significant evidence to resolve sequential cases. The latent prints are low quality, corrupted by noise and exhibit minor details. Image enhancement is compulsory in latent prints to transform the latent image into superior quality image. To rectify these issues, an automated latent fingerprint identification system is presented here with the aid of convolution neural network (CNN) of deep machine learning algorithm. The images are generally imperfect and complicated to categorize. Therefore, appropriate enhancement processes are made for pre-processing the fingerprint images; i.e., the minutiae features are extracted from the fingerprint images. These features are given to the CNN network as input for training as well as testing. The performance evaluation is done by calculating precision, recall, f1-score and accuracy. The experimental results are made by implementing in python where the proposed achieves a high accuracy rate of 99% recognition rate.