Robust approach for sensor placement in automated vision dimensional inspection

This paper introduces a novel sensor placement approach for dimensional inspection with active visual inspection systems. For a given set of entities, an effective sensor plan includes parameter values that allow some entities to be observed simultaneously. Two kinds of approaches have generally been used for sensor planning in visual inspection systems: the generate-and-test method and the synthesis method. In order to have high performance and satisfactory accuracy, uncertainty must be considered in the sensing model of the visual system. Furthermore, in order to avoid sharp deterioration of the performance of the visual inspection system for very small disturbances of the parameters, an appropriate definition of robustness is proposed. In an experiment a new sensor planning approach based on robustness resolves some problems with earlier approaches.