Discriminating fingermarks with evidential value for forensic comparison

Abstract

Law enforcement agencies all around the world are using biometrics and especially fingerprints to solve and fight crime. Often forensic experts are needed to record fingermarks at crime scenes and to ensure that those captured are of forensic value. In times of increased demand for forensic services, this process needs to be automated and streamlined as much as possible to improve efficiency and reduce workload. Hence, we investigate if the forensic evidential value (suitability for forensic analysis and/or examination) of fingermark images can be determined at an early stage automatically without any expert involvement, especially when using a mobile phone camera. We explore different factors such as the capture device and the constraints inferred, image feature sets and classifiers used, and their interplay. A database of 1;428 pseudo fingermarks has been collected and its ground truth, whether a mark is of forensic value or not, has been determined by 3 experts. The lowest equal error rate achieved, when using a mobile phone to capture the marks, is 13:62%. These promising results suggest that it might be possible to streamline forensic procedures by the application of an independent automated tool to assist with certain tasks.