A Point Cloud Registration Method for Antarctic Power Cabins Based on an Improved SAC-IA Algorithm

This paper presents a point cloud registration method for Antarctic power cabins based on an improved SAC-IA algorithm. To address the issues of measurement errors in structured light scanning and edge structure damage caused by traditional mean depth filtering, as well as inaccurate random point pair selection and long computation times in the SAC-IA algorithm, the following work was conducted: First, an improved depth filtering algorithm was designed to better filter out depth noise while preserving edge information. Second, ISS keypoint detection and cosine similarity were used to match the best feature pairs, combined with the RANSAC method to remove incorrect matches. Experimental results show that the improved algorithm significantly optimizes the number of key points and matching pairs, greatly saves computational resources, and enhances registration accuracy and robustness.