Automatic Latent Fingerprint Pose Estimation.

In this study, we propose a novel method towards automatic pose estimation in latent fingerprint images. Proposed method utilizes one of the most popular models in image classification, namely MobileNetV3, for pose estimation using latent images. With the pre-processing steps applied before classification, the region where the traces are found in the crime scene traces is segmented and the noise level is reduced. The proposed method was tested using 2034 tagged latent fingerprints obtained from the General Directorate of Security (EGM) in Ankara, Turkey. Proposed methods improves the classification accuracy considerably compared to an earlier method developed towards this task.