A Symmetry-based Validator and Refinement System for Pedestrian Detection in Far Infrared Images

This paper presents a refinement and validation stage for a tetra-vision based pedestrian detection system. The complete system is based on the use of both visible and far infrared cameras; in an initial phase it produces a list of areas of attention in the images which can contain pedestrians. Starting from the assumption that human shapes are mostly symmetrical and feature a number of vertical edges, the symmetry and edge density content of these areas is computed. The results is used to filter the list, discarding areas of attention that feature a scarce symmetry content. At the same time, the size of each area is refined to match the symmetry content and therefore to better adapt to human shapes. Moreover, this module can be successfully used to partition areas that contains two or more human shapes.