ROBUST MACHINE VISION BASED PARTS INSPECTION: INTELLIGENT NEURO-FUZZY VERSUS THRESHOLD BASED CLASSIFICATION

摘要

Past experience with an industrial machine vision-based parts inspection system highlighted the need for a robust system, that is a vision system that could adapt to changes in the operating environment without requiring excessive retuning of the data analysis algorithm. With this need in mind, an intelligent neuro-fuzzy based image processing algorithm was developed and tested against a traditional threshold based algorithm. Experimental results indicate that the intelligent algorithm performs well when the data is not well segmented.