Sensing And Recognition Of Rigid Objects Using Structured Light

Work directed toward the development of a vision system for bin picking rigid threedimensional (3-D) objects is reported. Any such system must have components for sensing, feature extraction, modeling, and matching. A structured light system for 3-D sensing and feature extraction, which attempts to deliver a rich 2–D representation of the scene, is described. Surface patches are evident as connected sets of stripes whose 3-D coordinates are computed via triangulation and constraint propagation across the stripe network. Object edges are detected via intersection of surface patches or by backprojecting image edges to intersect with patches in the 3-D space. Two matching paradigms are given for drawing correspondence between structures in the scene representation and structures in models. Three major contributions are contained in the work. First, original work is given for sensing object surface patches without having to uniquely solve for stripe labels. Second, use of both an intensity image and a striped image allows detected edges to be used along with 3-D surface patches to represent the scene. Finally, a pose clustering algorithm offers a uniform technique to accumulate matching evidence for recognition while, at the same time, averaging out substantial errors of pose computed from individual feature matches.