A Coarse-to-fine Approach for Rectifying Distorted Latent Fingerprints from Crime Scenes

With the widely used fingerprints to identify criminals, the influence of fingerprint deformation arouses attention in forensic science. To date, many approaches have been proposed to rectify distorted fingerprints. However, the performance in handling low-quality fingerprints extracted at a crime scene is less high quality than expected. This paper presents a combined method to rectify the latent fingerprints extracted at a crime scene. The method is a coarse-to-fine approach, combining the robustness of traditional pattern recognition and the accuracy of deep learning networks. We conducted several experiments to compare our approach with other approaches, including the nearest-neighbor search and network methods. The results show a remarkable improvement in fingerprint matching, especially in low-quality latent fingerprints. The top 25 cumulative match rate improves from 0.65 (original) to 1 (proposed method), whereas other approaches improve the result to 0.85 at best.