Footwear impressions are frequently found at various crime scenes. They are easily detected, processed and interpreted, and are therefore a valuable source of evidence for criminal investigations. Especially the combination with other types of forensic evidence e.g. DNA, toolmarks, fingerprints offers a great potential for solving a crime. Additionally to an estimate of the shoe size, the unique patterns of footwear impressions contain clues to the model and brand of the footwear, which in turn help to limit the number of possible suspects. Further, similar footwear impressions at different crime scenes indicate that the crime was committed by the same suspect. This way criminal acts committed by serial offenders can be identified. For instance burglaries are a great unease for society and are mostly committed by serial offenders. Solving those cases is a crucial factor in improving the subjective sense of security of the people.
In case a suspect is apprehended, the individual features of the footwear can proof that a specific shoe made a footwear impression. For this, forensic experts investigate the individual wear, damages and manufacturing marks. If multiple matching features can be found, the forensic evidence can support the prosecution in court. However, for this investigation the actual shoe has to be retrieved, from either the suspect or the evidence locker, and compared to the footwear impression. Since this process is time consuming and cumbersome, a limitation of the number of necessary comparisons to the most similar footwear impressions is desired by the forensic experts. Therefore, an automatic system that helps searching through databases with thousands of footwear impression images is needed. However, the currently used software solution is ill equipped to solve this problem. The main problem is that the footwear impressions have to be classified by the forensic expert by hand. This is done by describing the unique patterns of the impressions using a set of predefined classes. However, this process is very subjective and therefore the resulting list of similar impressions is not able to accurately depict the footwear impressions that were made by the same shoe. To alleviate this problem the goal of the impress project is an automated system, which implements an efficient image comparison methodology to find similar footwear impressions in huge databases of images.
Further, to allow an identification of the shoe model and brand a footwear impression reference database, i.e. shoe catalog, is created using the huge amount of shoe sole images freely available in the internet. Since the usage of those images is restricted, the legal framework is investigated in detail.
ProjektleiterIn / Name und Institut/Unternehmen:
Univ.Prof. DI Dr. Robert Sablatnig
Computer Vision Lab
Faculty of Informatics / Institute of Visual Computing & Human-Centered Technology
+43 1 58801 – 193161