Principal component analysis based image fusion routine with application to stamping split detection

This dissertation presents a novel thermal and visible image fusion system with application in online automotive stamping split detection. The thermal vision system scans temperature maps of highly reflective steel panels to locate abnormal temperature readings indicative of high local wrinkling pressure that causes metal splitting. The visible vision system offsets the blurring effect of thermal vision system caused by heat diffusion across the surface through conduction and heat loss to the surroundings through convection. The fusion of thermal and visible images combines two separate physical channels and provides more informative result image than either of the original ones.Principal Component Analysis (PCA) is employed to transform original image to its eigenspace. By retaining the principal components with influencing eigenvalues, PCA keeps the key features in the original image and reduces noise level. Then pixel level image fusion algorithms are developed to fuse original images from the thermal and visible channels, enhance the result image from low level and reduce undesirable noises. Finally, an automatic split detection algorithm is designed and implemented to perform online objective automotive stamping split detection.The integrated PCA based image fusion system for stamping split detection is developed and tested on an automotive press line. It is also assessed by online thermal and visible acquisitions and illustrates performance and success. Different splits with variant shape, size and amount are detected under actual operating conditions.