We are pleased to announce a new scientific publication from the CIRCULess project titled: “A Minimal Subset Approach for Informed Keyframe Sampling in Large-scale SLAM” led by the team at Luleå University of Technology.
📌This work addresses the growing computational burden of loop closure detection in large-scale LiDAR SLAM systems, where identifying and validating candidate loop closures becomes increasingly expensive as mission size scales. Although keyframe sampling is essential for managing this complexity, existing methods often rely on heuristic rules that fail to effectively balance redundancy reduction with information preservation.
To overcome this limitation, this work introduces an online keyframe sampling strategy that explicitly selects the most informative frames for global optimization. The proposed Minimal Subset Approach (MSA) operates in feature space within a sliding window framework, reducing redundant keyframes while retaining essential information for robust and efficient loop closure detection.
📄 Read the paper here: A Minimal Subset Approach for Informed Keyframe Sampling in Large-Scale SLAM | IEEE Journals & Magazine | IEEE Xplore and find all publications from CIRCULess here: https://circuless-project.eu/circuless/dissemination/#publications
Thank you to all contributors for their excellent work: Nikolaos Stathoulopoulos, Christoforos Kanellakis, George Nikolakopoulos
