Automatic Sensor Placement for Accurate Dimensional Inspection

摘要

Deriving accurate 3D object dimensions with a passive vision system demands, in general, the use of multistation sensor configurations. In such configurations, object features appear in images from multiple viewpoints, facilitating their measurement by means of optical triangulation. Previous efforts toward automatic sensor placement have been restricted to single sensor station solutions. In this paper we review photogrammetric expertise in the design of multistation configurations, including the bundle method—a general mathematical model for optical triangulation—and fundamental considerations and constraints influencing the placement of sensor stations. An overview of CONSENS, an expert system-based software tool which exploits these considerations and constraints in automatically designing multistation configurations, is given. Examples of multistation configurations designed by CONSENS demonstrate the tool's capabilities and the potential of our approach for automating sensor placement for inspection tasks.