Optimization fingerprint reconstruction using deep learning algorithm

Fingerprint recognition is one of the most well-known digital identifications and has been widely used on forensic science, criminal investigation, financial services, electronic smart locks...etc. In this paper, latent fingerprint marks have been image segmentation and reconstruction based on the Unet method. In the first, latent fingerprint marks were collected on Ninhydrin reaction thermal-induced paper and the image of fingerprints were segmented using Unet algorithm. Secondly, mutilated fingerprints were image reconstructed for the whole loops and whorls on a finger. And lastly, a Receiver Operating Characteristic (ROC) curves scheme has been applied to analyzed classification accuracy of a statistical developed model.