Remarks on a real-time 3D human body posture estimation methodusing trinocular images

摘要

This paper proposes a new real-time method of estimating humanpostures in 3D form trinocular images. The proposed method extractsfeature points of the human body by applying a type of function analysisto contours of human silhouettes. To overcome self-occlusion problems,dynamic compensation is carried out using the Kalman filter and allfeature points are tracked. The 3D coordinates of the feature points arereconstructed by considering the geometrical relationship between thethree cameras. Experimental results confirm both the feasibility and theeffectiveness of the proposed method, and an application example of the3D human body posture estimation to a motion recognition system ispresented