A fast image stitching algorithm in the endless hot rolling process

Camber is one of the most important defects in the endless hot-rolling process, and can be defined as the longitudinal curvature of a slab or strip in the plan view. To obtain a value of the camber for a steel bar, the overall shape of a slab or strip is required. However, the strip is too long to capture overall image at once. In order to obtain precise information on the overall shape of the bar, two area-scan charge-coupled device (CCD) cameras to obtain partial bar images and the image stitching algorithm are needed. This paper proposes a fast algorithm to obtain the overall shape of a steel bar in the endless hot rolling process. The proposed algorithm reduces the computational complexity and the running time of the image stitching process by adaptively selecting the search region to find feature points in steel bar images. The algorithm consists of three parts: generating the feature points and descriptors of the selected regions of continuous images for the steel bar, choosing the overlapped region between the present image and the next image, and stitching the partial bar images after calibrating them. The proposed stitching algorithm based on adaptive selection of the search region improves the performance of the camber measurement system during the image stitching process.