The work presented here is part of a project in collaboration with NASA-Ames to enhance Human-Computer Interaction by tracking the hand in space from multiple images. Hand pose estimation is a very challenging problem which can be decomposed into estimating the pose of the palm and the pose of the fingers. By placing elliptical markers on the back of the palm of the hand, we can accurately and efficiently recover the pose of the palm from several images.;This thesis addresses the multiple camera calibration problem, in addition to the palm pose estimation from the distortion of one or two coplanar ellipses in multiple-cameras.;Two existing ellipse pose estimation algorithms and a new multiple-camera model-based ellipse tracking algorithm are presented and implemented. All three algorithms are compared in our multiple-camera environment. As a possible application of our new algorithm, an extrinsic parameter recalibration technique is introduced.
展开▼
