[Japanese | Thesis | Researches in Minoh Lab | Minoh Lab]
In this paper, we propose a method to acquire the pose of an articulate object from succesive volume data. We use a structure model of an articulate object and no shape model of it. The pose of the articulate object is defined as the location of joint points and articular angles according to the structure model. We estimate the rotation center of two rigid regions as a joint point candidate (JPC) from two successive volume data, eliminate the false JPCs, and detect undetected JPCs by using successive volume data incrementally. Then we make correspondence between estimated JPCs and joint points of the structure model, and estimate the articular angles based on the estimated rigid regions and joint points.