Image Stitching Algorithm for Camber Measurement in Hot Rolling Process: Cross-Correlation Approach (ICCAS 2015)

Abstract

In the endless hot rolling process, defects should be controlled to improve quality and productivity. Camber is one of the most significant defects which cause decreasing of quality and productivity of steel. To handle the camber, the process that measures the camber exactly should be preceded. For measuring the camber of the strip, the whole image of strip is necessary. If the camera of the measurement system can only capture the part of the strip, we should stitch the consecutive partial images from the camera to obtain the whole image of strip. In this case, the partial consecutive images are consist of two parts which are overlapped region and new additional region. To stitch the partial images properly, we should know the length of new additional region. This paper proposes a novel algorithm to obtain the length of new additional region of each images based on the cross-correlation approach which has fast computational speed and high accuracy. Experimental results are given to demonstrate the effectiveness of the proposed algorithm.