Robust Image Registration for Power Equipment Using Large-Gap Fracture Contours

The registration of multimodal images is a fundamental and essential problem in various applications, including medical imaging, remote sensing, and computer vision. In computer vision, automatic registration of infrared and visible images of power equipment has become a crucial step in smart grid diagnostics. This article presents a robust image registration method for power equipment using large-gap fracture contours. The proposed approach employs the curvature scale space algorithm for feature detection, random sample consensus for robust matching, and best bin first for efficient search. The experimental results demonstrate improved registration accuracy and reduced computational time compared to existing methods.