DS-SLAM: A Semantic Visual SLAM towards Dynamic Environments. 作者：Chao Yu. 文献笔记
摘要
需要解决的问题：However, some problems are still not well solved, for example, how to tackle the moving objects in the dynamic environments, how to make the robots truly understand the surroundings and accomplish advanced tasks.
本文所做的工作：Five threads run in parallel in DS-SLAM: tracking, semantic segmentation, local mapping, loop closing and dense semantic map creation. DS-SLAM combines semantic segmentation network with moving consistency check method to reduce the impact of dynamic objects, and thus the localization accuracy is highly improved in dynamic environments.
本文结果优势：The results demonstrate the absolute trajectory accuracy in DS-SLAM can be improved one order of magnitude compared with ORB-SLAM2.
介绍
本文主要内容：1. A complete semantic SLAM system in dynamic environments (DS-SLAM) is proposed based on ORB-SLAM2 [2], which could reduce the influence of dynamic objects on pose estimation. The results indicate that DS-SLAM outperforms ORB-SLAM2 significantly regarding accuracy and robustness in dynamic environments.
2. We put a real-time semantic segmentation network in an independent thread, which combining semantic segmentation with moving consistency check method to filter out dynamic portions of the scene, like walking people.
3. DS-SLAM creates a separate thread to build a dense semantic 3D octo-tree map.
展望
- the types of objects that can be recognized in semantic segmentation network are restricted, which limit its scope of application.
- we would improve the performance of DS-SLAM in terms of real time by optimizing the moving consistency check method.
- the dense semantic octo-tree map would be adopted
for the mobile robots to accomplish high-level tasks.