Robot vision for automated manufacturing is a difficult and challenging problem. A robot vision system for assembly tasks must be able to analyze images including parts, tools, and the robot arm itself. Specular surfaces and textures are among the problems that will cause difficulty in segmenting the images. Our long term research objective is to construct an integrated robot vision system for automated manufacturing. We propose to first evaluate three promising image processing methodologies mathematical morphology, the facet model for image data, and the topographic primal sketch-on a representative set of images. We will then determine how the information extracted from these images can be used in an integrated system to extract the threedimensional information needed by the robot controller. We anticipate that the Phase I research will produce a significant study of low to mid-level vision techniques and the preliminary design of an integrated system. The Phase 11 research will select one or more specific assembly tasks, and a working system will be designed and constructed.The potential commercial application as described by the awardee: Potential commercial applications include all kinds of three dimensional robot guidance tasks in domains such as the automotive and aerospace industries. Specific examples include assembiv of automobile bodies, putting wheels on the axles of cars, application of sealant in car body seams, and assembly of appliances on missiles.
