Abstract

Footwear marks recovered from crime scenes are an important source of forensic intelligence or evidence - for some crime types, there is a greater probably to recover footwear marks than fingerprint ones. Currently the process of identifying a specific shoe model from the 10,000s of possibilities is a time-consuming task for expert examiners. As with many other crime marks, for example latent fingerprints, there is an increasing need for automation. The emergent research effort in this field has been hampered by the lack of a suitable dataset of footwear impressions. We present, here, a substantial and fully characterized dataset together with a proposed methodology for its use. © 2009 Springer Berlin Heidelberg.