ABSTRACT
This chapter introduces the laser surface heat treatment process and explains the methods used to acquire the image sequences. It describes the machine learning strategy applied to build the automated visual inspection (AVI) system capable of analyzing these image sequences and automatically performing in-process quality control. The chapter discusses the performance assessment of the AVI system. It also describes the spatio-temporal relationships of the laser process deduced from the machine learning model. The chapter summarizes the main conclusions and future directions. It reviews the main machine learning techniques used for building anomaly detection systems within image processing applications. The chapter explores the most extended techniques in the literature used for learning the normal behavior of temporal systems. It observes that the spatio-temporal relationships automatically learned with DBNs were consistent with the properties of the laser process in normal conditions, namely, the direction of the movement of the spot and the stability of the temperature.
