Loading...

Semantic Visual SLAM in Dynamic Environments

Habibpour, Mobin | 2022

228 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55365 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Meghdari, Ali; Nemati Estahbanati, Alireza; Taheri, Alireza
  7. Abstract:
  8. Most of the existing visual SLAM methods heavily rely on a static world assumption and easily fail in dynamic environments. One solution is to eliminate the influence of dynamic objects by introducing deep learning-based semantic information to SLAM systems. In this project, we propose a real-time semantic RGB-D SLAM (built upon RTAB-Map) system for dynamic environments that is capable of detecting moving objects and maintaining a static map for robust camera tracking. Furthermore, we augment the semantic segmentation process using an Extended Kalman filter module to detect temporarily static moving objects by adding centroids to each found dynamic object and calculating their velocity. We have also implemented a generative network to fill up the deleted regions of input images belonging to dynamic objects. Results have shown satisfactory filling performance but low fps from this module. This highly modular framework has been implemented on the ROS platform and can achieve around 17 fps on a GTX 1080. Benchmarking the developed pipeline on dynamic sequences from the TUM dataset suggests that the approach achieves far lower localization error compared to baseline SLAM, while also providing the position of the tracked dynamic objects, a 3D map free of those dynamic objects, better loop closure detection with the whole pipeline able to run on a robot moving at moderate speed.
  9. Keywords:
  10. Semantic Segmentation ; Extended Kalman Filter ; Mobile Robot ; Dynamic Enviroment ; Generative Networks ; Simultaneous Localization and Mapping (SLAM)

 Digital Object List

 Bookmark