Abstract

In this paper, a deep learning-based semantic SLAM (simultaneous localization and mapping) algorithm for dynamic environment is proposed to address the problem that the presence of dynamic objects in the environment has an impact on map building. First, the semantic segmentation network is used to semantically segment the image and obtain the semantic information of the objects in the environment; second, the semantic information is incorporated into the visual odometry and the feature points are tracked using the optical flow method to detect dynamic objects and remove them from the map building process; finally, the two-dimensional semantic information generated by the semantic segmentation network is fused with the three-dimensional geometric information generated by the visual SLAM system to semantic maps are constructed and validated on the TUM dataset, confirming that the algorithm can build static semantic maps without moving objects in dynamic environments.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call