A0983
Title: Monitoring time dependent image processes
Authors: Yarema Okhrin - University of Augsburg (Germany) [presenting]
Abstract: Monitoring techniques are developed for high dimensional image processes with temporal dependence. This problem is of great practical importance in visual quality control of production processes. It is assumed that the pixel intensities follow a spatial AR process. In the out-of-control state we expect a location shift. To overcome computational infeasibility of the implementation we focus on dimension reduction techniques based on averaging. Moreover, we derive necessary and sufficient conditions for the original process and spatially averaged process to follow the same spatial AR process. For monitoring purposes we apply several multivariate EWMA control charts and evaluate their performance in extensive simulations.
