Surface descriptions from vision and touch

The goal of vision is object recognition. Recent research has shown that the human visual system creates a surface description of a scene, including depth and orientation information at all points in a scene as a first step before creating an object centered description. This description is referred to as a 2 1/2 D sketch. Machine vision systems presently do not have the capability of creating this 2 1/2 D sketch from visual information alone, especially for curved surface objects. By using tactile data in cooperation with vision, a method is proposed for creating a surfacc description of an object. This surface description uses bicubic surface patches as a primitive. Once a surface sketch is created with bicubic surface patches the next steps in the hierarchy of processing arc feasible, including a transformation to a full 3D object centered description.