On the reconstruction of 3D shape using wireframes

Describes a new algorithm for the reconstruction of 3D shape from passive scene analysis, using a novel approach based on region growing, and polygon fitting of shapes detected using up to three cameras i.e. from a three camera sensor head. The visual cues used for matching between different views are the vertices of the polygon fits to each region. The matching is achieved using 3D reconstruction geometry based on the known position and orientation of each camera in the world coordinate frame. The method has been called the inverse perspective method. A prototype system has been built at Newcastle using transputers, and initial experimentation has been encouraging. Wireframe generation is benchmarked at 2.5 s per update