Globally Optimal and Scalable Video Image Stitching for Robotic Visual Inspection of Electric Generators

Large-scale image stitching of temporal video data acquired by a small robot can facilitate the inspection process of electric generators, reduce the inspection time, and improve the reliability. However, the image data poses a number of challenges due to the small field of view, lack of distinct texture, specular highlights, and other image artifacts. We introduce a novel image stitching method, which generates composite images of generator wedges from temporal videos using intensity-based registration and non-linear blending. In contrast to previous intensity-based registration approaches, our global method simultaneously exploits the information of all image frames of a video and directly determines the global image translations. We propose a suitable energy function and employ a graph-based method for globally optimal minimization in linear runtime. Regularization is used to exploit physical knowledge about the application domain which improves the robustness. We have applied our approach to temporal video data of rotor wedges and performed a comparison with previous methods. We found that our method yields superior results.