This project is the result of my undergraduate dissertation. The localization in a dynamic environment is to deploy TensorRT optimized YOLOX in the front end of ORB-SLAM3 for object detection and eliminate all points belonging to the human bounding box. At the same time, the semantic information is sent to the mapping module to dye the 3D point cloud. The disadvantage of this project is that in the localization module, only the points belonging to people are processed because people are dynamic most of the time. In the mapping module, we did not segment semantic objects accurately, resulting in the wrong coloring of point clouds of other objects.
The main part of the code will be published after the paper based on this is received by publication.
The video is presented below:
Other countries:https://www.youtube.com/watch?v=OxYHrIgqyJQ
Our work is based on ORBSLAM3,Crowd-SLAM,Semantic-SLAM
@article{ORBSLAM3_TRO,
title={{ORB-SLAM3}: An Accurate Open-Source Library for Visual, Visual-Inertial
and Multi-Map {SLAM}},
author={Campos, Carlos AND Elvira, Richard AND G\´omez, Juan J. AND Montiel,
Jos\'e M. M. AND Tard\'os, Juan D.},
journal={IEEE Transactions on Robotics},
volume={37},
number={6},
pages={1874-1890},
year={2021}
}
@article{soaresJINT2021,
title={Crowd-{SLAM}: Visual {SLAM} Towards Crowded Environments using Object Detection},
author={Soares, J. C. V., Gattass, M. and Meggiolaro, M. A.},
journal={Journal of Intelligent & Robotic Systems},
volume={102},
number={50},
doi = {https://doi.org/10.1007/s10846-021-01414-1},
year={2021}
}
@inbook{inbook,
author = {Qi, Xuxiang and Yang, Shaowu and Yan, Yuejin},
year = {2018},
month = {10},
pages = {012023},
title = {Deep Learning Based Semantic Labelling of 3D Point Cloud in Visual SLAM},
volume = {428},
journal = {IOP Conference Series: Materials Science and Engineering},
doi = {10.1088/1757-899X/428/1/012023}
}
Please refer to the compilation and operation of ORB-SLAM3:https://github.com/UZ-SLAMLab/ORB_SLAM3