Accurate, high speed, and robust inspection enabled by salient landmarks

Quality control of prints is an important industrial application of machine vision. Commercial vision libraries provide different tools (e.g. barcode recognition, template matching) that facilitate the development of complete inspection systems. However, closed source software is usually not compatible with the modification of functions for optimization purposes, hampering the solution of specific problems with state-of-the-art approaches. In this paper, we propose the use of a simple measure of saliency to extract small landmarks to speed up a matching-based alignment. We show the simplicity and efficacy of the adopted approach, as well as the compatibility with any prototyping environment with minimal functions for image manipulation. It enables a fast prototyping of a robust inspection solution at a very low cost and requiring a minimal intervention in a production line.