PV-SLAM: Panoptic Visual SLAM with Loop Closure and Online Bundle Adjustment


Ashok Bandyopadhyay (Indian Institute of Technology, Guwahati), Pranjal Baranwal (Indian Institute of Technology, Guwahati, Indian institute of science, Bangalore), Arijit Sur (Indian Institute of Technology, Guwahati), Rajeev UP (Vikram Sarabhai Space Centre, Indian Space Research Organization, Thiruvananthapuram, India)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)

Abstract

Visual Simultaneous Localization and Mapping (vSLAM), a variant of SLAM systems, is a navigation technique in which autonomous robots leverage visual data from camera to create a map of an unknown environment while concurrently determining their own position on that map. Panoptic segmentation-based Visual SLAM offers a more comprehensive, efficient, and robust solution for robot perception and scene understanding compared to approaches that rely solely on semantic or geometric information. In Visual SLAM, the panoptic segmentation simultaneously identifies and delineates all objects in an image, providing a detailed understanding of both their instances and semantic categories. It offers the potential to improve the robustness and accuracy of SLAM systems in complex and dynamic environments. Recent Panoptic segmentation based works on Visual SLAM ignore the problem of loop closure. Also, these methods use offline bundle adjustment, which can lead to drift errors, highlighting the need for online bundle adjustment. In this paper, we introduce a novel architecture that integrates loop closure and online bundle adjustment, expanding on the PVO(Panoptic Visual Odometry) model. The results show that our model outperforms state-of-the-art methods in visual odometry tasks. The ablation study shows that our technique outperforms current state-of-the-art methods, exhibiting superior performance across majority of sequences.

Citation

@inproceedings{Bandyopadhyay_2024_BMVC,
author    = {Ashok Bandyopadhyay and Pranjal Baranwal and Arijit Sur and Rajeev UP},
title     = {PV-SLAM: Panoptic Visual SLAM with Loop Closure and Online Bundle Adjustment},
booktitle = {35th British Machine Vision Conference 2024, {BMVC} 2024, Glasgow, UK, November 25-28, 2024},
publisher = {BMVA},
year      = {2024},
url       = {https://papers.bmvc2024.org/0936.pdf}
}


Copyright © 2024 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection