Development of a novel image sensor and its application to analysis of automobile components

摘要

The authors have developed a novel image sensor for continuous conditioned monitoring of high-precision tolerances of a complex automotive product. A novel mechanooptical arrangement has been designed and validated to capture the images/silhouettes of the components as input into a neural network designed on approximate reasoning architecture. The design is extensible to handle a large number of rules, and the speed of inference is almost independent of the number of rules.