Research on Intelligent Vehicle Traffic Scenes Mapping Algorithm Considering Dynamic Objects

Tongtong Liu,Honghui Dong,Yue Zhou,Huipeng Zhang,Wei Li,Qi Zhang
DOI: https://doi.org/10.1109/itsc57777.2023.10422535
2023-01-01
Abstract:The intelligent vehicle perception algorithm is a fundamental technology for constructing the intelligent transportation system. To better estimate the trajectory of intelligent vehicles and achieve traffic scene mapping, this paper proposes an intelligent vehicle mapping algorithm framework considering dynamic objects. The proposed framework can be applied to monocular, stereo and RGB-D cameras. It combines lightweight dynamic object detection algorithms (Yolov5s) with traditional visual simultaneous localization and mapping algorithms (ORB-SLAM2) to reduce the impact of dynamic objects on mapping accuracy. The algorithm uses an improved optical flow method combined with the Yolov5 object detection algorithm. Yolov5 is utilized to detect potential dynamic objects in traffic scenes, while the improved optical flow method filters out real dynamic objects containing dynamic feature points, effectively removing dynamic objects from the images captured by the intelligent vehicle-mounted camera. Experiments on the KITTI dataset show that our algorithm can effectively reduce the impact of dynamic objects on pose estimation in intelligent vehicle traffic scenes and reduce trajectory error compared to the ORB-SLAM2 algorithm. Additionally, the algorithm has advantages over typical dynamic SLAM algorithms in terms of realtime performance.
What problem does this paper attempt to address?