Automated Architectural Acquisition From A Camera Undergoing Planar Motion

Much recent research on structure and motion recovery has concentrated on the case of reconstruction of unstructured environments from an uncalibrated video sequence. Here we show the advantages that accrue to both the motion determination and structure recovery if constraints are available on the motion and environment. We consider the case of a camera with fixed internal parameters undergoing planar motion in an indoor environment for which several dominant directions occur. The novelty of this work is that it is shown that under these constraints the problems of both motion determination and 3D structure recovery can be reduced to a sequence of one parameter searches. This low dimensional search enables efficient, robust and reliable algorithms to be engineered. The resulting algorithms are demonstrated on images of very visually impoverished scenes, and the results are compared to ground truth.