A 3d Thermal Model For Real-Time Condition Monitoring Of Electrochemical Processes

This work aims to present an automated defect diagnosis system that fuses physical and geometrical features in order to construct a detailed 3D thermal model for industrial shop-floor monitoring. To accomplish that, a geometric camera calibration framework has been followed where the calibration accuracy of a custom-made calibration pattern with high contrast in the infrared modality has been investigated. Defections are then identified using an MRF-based probabilistic superpixel segmentation algorithm in real-time. The results of the proposed procedure are encouraging; with F1-measure reaching 91% in certain datasets containing data from electrochemical industrial processes thus further demonstrating the effectiveness of the proposed system.