Abstract
This paper proposes a human body posture estimation method using a 3D articulated human body CG model. The estimation method is based on 2D matching between the human silhouettes extracted from camera images and the model silhouettes projected onto corresponded camera planes. The candidates of the human model silhouette is generated by using Monte Carlo filter and the XOR distance is introduced to calculate the likelihood rate between silhouettes. Experimental results confirmed the feasibility of the estimation method.
