Model-based image analysis for forensic shoe print recognition

摘要

This thesis is about automated forensic shoe print recognition. Recognizing a shoe printudin an image is an inherently difficult task. Shoe prints vary in their pose, shape andudappearance. They are surrounded and partially occluded by other objects and mayudbe left on a wide range of diverse surfaces. We propose to formulate this task in audmodel-based image analysis framework.udOur framework is based on the Active Basis Model. A shoe print is represented asudhierarchical composition of basis filters. The individual filters encode local informationudabout the geometry and appearance of the shoe print pattern. The hierarchical com-udposition encodes mid- and long-range geometric properties of the object. A statisticaluddistribution is imposed on the parameters of this representation, in order to account forudthe variation in a shoe print‘s geometry and appearance.udOur work extends the Active Basis Model in various ways, in order to make it robustlyudapplicable to the analysis of shoe print images. We propose an algorithm that automat-udically infers an efficient hierarchical dependency structure between the basis filters. Theudlearned hierarchical dependencies are beneficial for our further extensions, while at theudsame time permitting an efficient optimization process. We introduce an occlusion modeludand propose to leverage the hierarchical dependencies to integrate contextual informa-udtion efficiently into the reasoning process about occlusions. Finally, we study the effectudof the basis filter on the discrimination of the object from the background. In this con-udtext, we highlight the role of the hierarchical model structure in terms of combining theudlocally ambiguous filter response into a sophisticated discriminator.udThe main contribution of this work is a model-based image analysis framework whichudrepresents a planar object‘s variation in shape and appearance, it‘s partial occlusion asudwell as background clutter. The model parameters are optimized jointly in an efficientudoptimization scheme. Our extensions to the Active Basis Model lead to an improveduddiscriminative ability and permit coherent occlusions and hierarchical deformations. Theudexperimental results demonstrate a new state of the art performance at the task ofudforensic shoe print recognition.