Shape and nonrigid motion estimation through physics-based synthesis

摘要: A physics-based framework for 3-D shape and nonrigid motion estimation real-time computer vision systems is presented. The features dynamic models that incorporate the mechanical principles of rigid bodies into conventional geometric primitives. Through efficient numerical simulation Lagrange equations motion, can synthesize physically correct behaviors in response to applied forces imposed constraints. Applying continuous Kalman filtering theory, a recursive estimator employs as system model developed. continually synthesizes generalized arise from inconsistency between incoming observations estimated state. observation also account formally instantaneous uncertainties incomplete information. Riccati procedure updates covariance matrix transforms accordance with dynamics prior history. Experiments involving fitting tracking articulated flexible objects noisy data are described. >