Pose Estimation and 3-D Modeling from Video by Volume Feedback

摘要

Volume reconstruction and pose retrieval of an arbitrary rigid object from monocular video sequences is addressed. Initially, the object pose is estimated in each image by locating similar textures, assuming a flat depth map. Then shape-from-silhouette is used to make a volume (3-D model). This volume is used in a new round of pose estimations, this time by a model-based method that gives better estimates. Before repeating this process by building a new volume, pose estimates are adjusted to reduce error by maximizing a novel quality measure for shape-from-silhouette volume reconstruction. The feedback loop is terminated when pose estimates do not change much, as compared to those produced by the previous iteration. Based on the theoretical study of the proposed system, a test of convergence to a given set of poses is devised. Reliable performance of the system is also proved by several experiments. No model is assumed for the object. Feature points are neither detected nor tracked, as there is no problematic feature matching or correspondence. Our method can be also applied to 3-D object tracking in video.