Production line quality control using infrared imaging

This paper addresses the problem how to bring advanced data analysis techniques to the reality of a production line in order to increase the productivity and cost-effectiveness while reducing failure rates and increasing reliability of the final product. The main goal was to develop techniques of fast thermal inspection for production line quality control using a knowledge-based machine vision system. The paper contains a description of the system as well as a proposition of the algorithm for automatic classification of devices on the base of information included in their infrared images. Data-driven pattern recognition, infrared imaging, and principal component analysis (PCA) were put together and resulted in a very effective production line quality control system. The algorithm has been validated using real production line data. Experiments revealed some interesting features of the proposed method, e.g. resistance to changes of the ambient temperature and early classification during the thermal transient state.