计算机视觉
计算机科学
人工智能
同时定位和映射
雷达跟踪器
噪音(视频)
特征(语言学)
目标检测
稳健性(进化)
夹持器
移动机器人
作者
Shuang Song,Qingyuan Zhu,Huosheng Hu
标识
DOI:10.1109/jsen.2025.3609911
摘要
LiDAR-based simultaneous localization and mapping (SLAM) methods that rely solely on geometric features often fail in degraded environments with sparse or ambiguous structures. We present RTS-SLAM, a visually semantic–enhanced, tightly coupled SLAM framework that integrates semantic point cloud information with principal component analysis (PCA) and Kalman filtering to provide robust priors for dynamic object detection. A visual consistency (VC) subsystem removes dynamic points, while a loop closure module and improved generalized iterative closest point (IGICP)-based LiDAR–inertial mapping and optimization (LIMO) enhance global consistency. By sharing semantic information across modules, RTS-SLAM achieves stable, dense mapping in weakly structured scenes. Experiments on the KITTI dataset and real-world vehicle tests demonstrate competitive performance against state-of-the-art methods.
科研通智能强力驱动
Strongly Powered by AbleSci AI