Robust manufacturing inspection and classification with machine vision

This paper presents and compares methods for inspecting products using machine vision systems. This research compares the Mahalanobis Taguchi System (MTS) with a method based on principal component transformation and multi-modal overlap methods, and which is called the Principal component Feature overlap Measure (PFM). In an example application, the PFM achieves significantly higher Signal/Noise-ratios (+80 dB) and equal or better classification performance than MTS with a lower number of classification features (-75%).