Bogotá - Ciencias - Maestría en Ciencias - Estadística · 2026
Image-based process monitoring : robust tensorial data approaches
Record only. The repository marks this work as “Acceso abierto”: the PDF is not public, so there is no text to read here. View the record in the repository.
Abstract
This work presents a robust Phase I monitoring methodology for image-based quality control, supported by tensor representations and a joint scheme with a T2 chart for features and a Q chart for residuals, calibrated to a fixed joint Type I error probability. Two robustification strategies are proposed: robustifying the decomposition using MacroPARAFAC and robustifying the T2 chart using robust MCD estimation. The performance of the proposed methods is assessed through a simulation study using the signal probability as the evaluation measure. Two real data applications based on the MVTec AD (bottle) dataset are also included to illustrate the practical behavior of the proposed control charts. (Texto tomado de la fuente)