三次元人体CGモデルを用いた多眼視マーカレスモーションキャプチャ(機械力学,計測,自動制御)

抄録

This paper proposes markerless human motion capture using a 3D articulated CG model of the human body. The method of estimating human body posture is based on 2D matching between human silhouettes extracted from camera images and model silhouettes projected onto the corresponding camera planes in virtual space. Candidates for human model silhouette are generated using a Monte Carlo filter-based algorithm, and the normalised core-weighted XOR distance is introduced to calculate the likelihood rate between silhouettes. Experimental results show the feasibility and effectiveness of the proposed method for achieving markerless human motion capture.