Real time image stitching using network synchronized machine vision cameras

Abstract

Steel rolling processes are classified into hot and cold rolling. Hot rolling involves heating a steel slab and then flattening it into thinner sections of strips. On the contrary, cold rolling involves stretching and rolling the steel sheets at room temperatures. The key performance indicator for both these processes is the surface quality of the steel sheets, which is evaluated by automatically inspecting the steel surfaces using high-speed machine-vision cameras. In such high-speed industrial machine-vision (MV) applications, the problem of field-of-view (FOV) expansion in real-time is most challenging, that are generally resolved by employing multiple high-resolution cameras similar to 360-degree camera systems and distributed aperture systems (DAS). In real-time, industrial applications such as surface inspection of steel sheets, FOV expansion problems can be resolved by synchronously capturing and stitching multiple images with overlapping regions to yield a seamless high-resolution stitched image. newline