Search papers, labs, and topics across Lattice.
The paper introduces LST-SLAM, a stereo thermal SLAM system designed for robust performance in kilometer-scale dynamic environments by addressing challenges in feature extraction, motion tracking, and pose optimization specific to thermal imagery. LST-SLAM employs self-supervised thermal feature learning, stereo dual-level motion tracking, and geometric pose optimization, enhanced by a semantic-geometric hybrid constraint to filter dynamic features. Experimental results on large-scale thermal datasets demonstrate that LST-SLAM outperforms existing SLAM systems like AirSLAM and DROID-SLAM in terms of robustness and accuracy.
Achieve robust kilometer-scale SLAM in dynamic thermal environments by combining learned thermal features with geometric and semantic constraints, outperforming state-of-the-art visual SLAM methods.
Thermal cameras offer strong potential for robot perception under challenging illumination and weather conditions. However, thermal Simultaneous Localization and Mapping (SLAM) remains difficult due to unreliable feature extraction, unstable motion tracking, and inconsistent global pose and map construction, particularly in dynamic large-scale outdoor environments. To address these challenges, we propose LST-SLAM, a novel large-scale stereo thermal SLAM system that achieves robust performance in complex, dynamic scenes. Our approach combines self-supervised thermal feature learning, stereo dual-level motion tracking, and geometric pose optimization. We also introduce a semantic-geometric hybrid constraint that suppresses potentially dynamic features lacking strong inter-frame geometric consistency. Furthermore, we develop an online incremental bag-of-words model for loop closure detection, coupled with global pose optimization to mitigate accumulated drift. Extensive experiments on kilometer-scale dynamic thermal datasets show that LST-SLAM significantly outperforms recent representative SLAM systems, including AirSLAM and DROID-SLAM, in both robustness and accuracy.
