No abstract

No abstract

No abstract

Application of non-invasive scan technologies for acquisition of latent fingerprints promise a better support of forensic and dactyloscopic experts when securing evidence at crime scenes. Furthermore, non-destructive acquisition preserve the chance of subsequent chemical and forensic analysis of left residue. Based on results of an ongoing research project with sensor industry partners, this paper presents a collection of 28 statistical, gradient-, and spectral densitybased features for latent fingerprint detection using low resolution scans. Within this work a chromatic white light (CWL) sensor is used for image acquisition. Furthermore, based on concepts of biometric fusion, a taxonomy for possible fusion strategies is presented and very first results for three different strategies on decision level are discussed. Experimental evaluation is performed based on scans of 1680 latent fingerprints on three different surfaces. The results show very good performance on planar, non-absorbing surfaces with uniform reflection characteristics with an detection rate of 2.51% in the best case. On the other hand difficulties are arising from surfaces with non-uniform/predictable reflection characteristics.

In forensic investigations, the recovering of latent fingerprints is one of the most essential issues. Driven by human experts, today this process is very time consuming. An automation of both examination of suspicious areas and acquisition of fingerprints lead on the one hand to the covering of larger surfaces and on the other hand to significant speed up of the evidence collection. This work presents an experimental study on capabilities of chromatic white-light sensor (CWL) regarding the contact-less localization of latent fingerprints on differently challenging substrates. The fully automatic CWL-based system is implemented from the acquisition through the feature extraction right up to the classification. The key objective of the work is to develop a methodological approach for the quantitative evaluation of the localization success. Based on the proposed performance measures, the optimal system parameters such as scan resolution, extracted features and classification scheme are specified dependent on the surface material. Our experiments from an actual project with the sensor industry partner show convincing localization performance on easy-to-localize and adequate performance on moderate-tolocalize substrates. The hard-to-localize substrates require further improvements of the localization system.