Multicues 3D Monocular Upper Body Tracking Using Constrained Belief Propagation

This paper describes a method for articulated 3D upper body tracking in monocular scenes using a graphical model to represent an articulated body structure. Belief propagation on factor graphs is used to compute the marginal probabilities of limbs. The body model is a loose-limbed model including attraction factors between adjacent limbs and constraints to reject poses resulting in collisions. To solve ambiguities resulting from monocular view, robust contour and colour based cues are extracted from the images. Moreover, a set of constraints on the model articulations is implemented according to human pose capabilities. Quantitative and qualitative results illustrate the efficiency of the proposed algorithm. Figure 1: Upper body tracking. First row: original image, front, right side and top views of the obtained limbs positions with a single camera. Second row: background subtraction, contours, face colour map and energy motion distance map.