A three-dimensional vision system is indispensable for human-like robots and intelligent industrial robots. In order to reconstruct three dimensional shape of an object, an eye system is developed by combining stereo cameras and a laser stripe system based on the fact that two systems are complementary to each other. Pattern recognition algorithms for processing and fusion of sensor data are proposed. Edges and corner points (vertexes) of an object are extracted from stereo images, and surfaces constructing the object and edge points in intersections between neighbor surfaces are found from laser stripe images. Finally, each processed image is fused together for extracting three dimensional surface description. The proposed algorithms have been experimentally validated for polyhedral objects.
展开▼