Quality-Driven Feature Adaptation for Robust Visual SLAM in Challenging Environments

Ziwei Ma, Di He, and Yiqian Li

Peer Reviewed

Abstract:	Visual Simultaneous Localization and Mapping (SLAM) is a foundational technology for autonomous systems, but its performance is often hindered by varying environmental conditions such as lighting changes, image blur, and dynamic scenes. To enhance the robustness of visual SLAM under such challenging environments, this paper proposes a novel approach that evaluates image quality across different regions of each image frame and computes confidence masks for feature points. By combining multiple image quality metrics, a weighted score is assigned to each feature point. These weights are integrated into the pose estimation process to adaptively prioritize reliable feature points. Extensive experiments on the EuRoC and ICL-NUIM datasets demonstrate that the proposed method significantly improves pose accuracy, particularly in sequences with difficult environmental conditions, offering a promising solution for enhancing the reliability and robustness of visual SLAM systems.
Published in:	Proceedings of the 38th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2025) September 8 - 12, 2025 Hilton Baltimore Inner Harbor Baltimore, Maryland
Pages:	2043 - 2054
Cite this article:	Ma, Ziwei, He, Di, Li, Yiqian, "Quality-Driven Feature Adaptation for Robust Visual SLAM in Challenging Environments," Proceedings of the 38th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2025), Baltimore, Maryland, September 2025, pp. 2043-2054. https://doi.org/10.33012/2025.20452
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In