A model-based viewpoint determination method is developed to estimate multiple object 3-D motion from an image sequence. Unlike other 3-D motion estimation methods, the motion is not calculated from optical flow or image feature displacement between image frames. Instead, object 3-D positions and orientations are obtained by determining object viewpoints for each image frame. There are two stages in the viewpoint determination. First, the relational structures in an edge line image are matched with the relational structures in 2-D projections of 3-D model lines to select an object model from the model library. In the second stage, the selected object model is verified using a viewpoint consistence constraint. An application of this method to real-world image sequences is shown.
展开▼