| Popis: |
The matching of the latent print obtained at crime scene with the stored database at law enforcement agencies is the most important forensic application. The performance of an automated latent fingerprint matcher is limited by the unwanted appearance or poor quality of the latent prints. This reason necessitates latent fingerprint investigators for feature markups and quality value determination. However, the reliability and consistency of the manual assessment are significantly affected by various factors involved in the forensic examination. This paper proposed an algorithm to determine latent fingerprint image quality through feature extraction followed by k-means classifier. The feature vector consists of the ridge clarity values, number of extracted minutiae, average quality of the minutiae, area of the convex hull including all minutiae and textural feature values. Experimental results show that the identification rate of the minutiae-based latent fingerprint matcher is improved after rejecting unacceptable quality of query latent prints. |
