Multi-part Non-rigid Object Tracking Based on Time Model-Space Gradients

This paper presents a shape and pose estimation method for 3D multi-part objects, the purpose of which is to easily map objects from the real world into virtual environments. In general, complex 3D multi-part objects cause undesired self-occlusion and non-rigid motion. To deal with the problem, we assume the following constraints: object model is represented in a tree structure consisting of deformable parts.connected parts are articulated at one point (called “articulation point”).as a 3D parametric model of the parts, we employ deformable superquadrics (we call DSQ). To estimate the parameters from the sensory data, we use time model-space gradient method, which reduces the parameter estimation problem into solving a simultaneous linear equation. We have demonstrated that our system works well for multiple-part objects using real image data.