Machine Vision Based Approach for Automated Keypad Inspection

Abstract

Humans have been traditionally tasked with quality control because it requires judgment. Hence when it comes to manufacturing and quality inspection, Machine Vision plays an important role. However, for the best quality, parameters like lighting, changes in colour, curvature, alignment, required time for flaw detection are various challenging factors. Defects on flat products like keypad labels can be hard for human inspectors to spot easily, reliably due to their random occurrence. It is very hard to scrutinize keypads with a very high degree of detail and different colours within a small area. Hence for the inspection of flaws or even small defects in the keypad, an automated system is required which will select the correct, non-defective samples and reject the defective samples. While working with the images for detection of faulty regions, image subtraction is to be carried out. This is possible if two objects are partially aligned. Hence the need for image registration arises. The main aim of this paper is to register multi-temporal images for minimizing the misalignment of the objects and computation of high correlation factor by Hough Transform based Parameter Extraction (HTPE) and Image Registration using Multilevel Pyramids (IRMP). In the presence of non-uniform lighting conditions the accuracy achieved is 75%.