Shoeprints are one of the most commonly found evidences at crime scenes. A latent shoeprint is a photograph of the impressions made by a shoe on the surface of its contact. Latent shoeprints can be used for identification of suspects in a forensic case by narrowing down the search space. This is done by elimination of the type of shoe, by matching it against a set of known shoeprints (captured impressions of many different types of shoes on a chemical surface). Manual identification is laborious and hence the domain seeks automated methods. The critical step in automatic shoeprint identification is Shoeprint Extraction - defined as the problem of isolating the shoeprint foreground (impressions of the shoe) from the remaining elements (background and noise). We formulate this problem as a labeling problem as that of labeling different regions of a latent image as foreground (shoeprint) and background. The matching of these extracted shoeprints to the known prints largely depends on the quality of the extracted shoeprint from latent print. The labeling problem is naturally formulated as a machine learning task and in this paper we present an approach using Conditional Random Fields(CRFs) to solve this problem. The model exploits the inherent long range dependencies that exist in the latentprint and hence is more robust than approaches using neural networks and other binarization algorithms. A dataset comprising of 45 shoeprint images was carefully prepared to represent typical latent shoeprint images. Experimental results on this data set are promising and support our claims above.
展开▼
