Quality inspection of food packaging seals using machine vision with texture Tanalysis

This paper presents a machine vision approach to inspecting the integrity of semirigid polymer food packages sealed using a new laser-based sealing system. This new technique for food package sealing seeks to make significant improvements to existing manufacturing methods to meet the industryʹs requirements for rapid response to retail customers, while maintaining high quality, through 100 per cent inspection, with low associated production costs. By analysing examples of ʹgoodʹ and ʹbadʹ seal images, seal uniformity has been examined and associated with numerical quality measures. Statistical texture analysis is utilized in order to distinguish between good and bad seal conditions. A minimum distance classifier and an artificial neural network have been used to accept the segmented texture parameters as inputs and to output the seal quality decision. The method is shown to have a high success rate provided that illumination conditions remain constant.