Drift-Free Tracking of Rigid and Articulated Objects

Abstract

Model-based 3D tracker estimate the position, rotation, and joint angles of a \ngiven model from video data of one or multiple cameras. They often rely on \nimage features that are tracked over time but the accumulation of small errors \nresults in a drift away from the target object.\nIn this work, we address the drift problem for the challenging task of human \nmotion capture and tracking in the presence of multiple moving objects where \nthe error accumulation becomes even more problematic due to occlusions. To this \nend, we propose an analysis-by-synthesis framework for articulated models. It \ncombines the complementary concepts of patch-based and region-based matching to \ntrack both structured and homogeneous body parts. The performance of our method \nis demonstrated for rigid bodies, body parts, and full human bodies where the \nsequences contain fast movements, self-occlusions, multiple moving objects, and \nclutter. We also provide a quantitative error analysis and comparison with \nother model-based approaches.