An Intensity-Enhanced LiDAR SLAM for Unstructured Environments

ZHiqiang Dai,Jingyi Zhou,Tianci Li,Hexiong Yao,Shihai Sun,Xiangwei Zhu
DOI: https://doi.org/10.1088/1361-6501/acf38d
IF: 2.398
2023-08-25
Measurement Science and Technology
Abstract:Traditional LiDAR Simultaneous Localization And Mapping (SLAM) methods rely on geometric features such as lines and planes to estimate pose. However, in unstructured environments where geometric features are sparse or absent, point cloud registration may fail, resulting in decreased mapping and localization accuracy of the LiDAR SLAM system. To overcome this challenge, we propose a comprehensive LiDAR SLAM framework that leverages both geometric and intensity information, specifically tailored for unstructured environments. Firstly, we adaptively extract intensity features and construct intensity constraints based on degradation detection, and then propose a multi-resolution intensity map construction method. The experimental results show that our method achieves a 55% accuracy improvement over the pure geometric LiDAR SLAM system and exhibits superior anti-interference capability in urban corner scenarios. Compared with Intensity-SLAM, the advanced intensity-assisted LiDAR SLAM, our method achieves higher accuracy and efficiency.
engineering, multidisciplinary,instruments & instrumentation
What problem does this paper attempt to address?