REDUAS 2019 Paper Abstract

Close

Paper TuD22T2.1

Qian, Huaming (Harbin Engineering University), Ding, Peng (College of Automation, Harbin Engineering University)

An Improved ORB-SLAM2 in Dynamic Scene with Instance Segmentation

Scheduled for presentation during the Regular Session "Navigation" (TuD22T2), Tuesday, November 26, 2019, 10:10−10:30, Room T2

2019 Workshop on Research, Education and Development of Unmanned Aerial Systems (RED UAS), November 25-27, 2019, Cranfield University, Cranfield, UK

This information is tentative and subject to change. Compiled on January 20, 2022

Keywords Navigation

Abstract

In order to improve the accuracy of ORB-SLAM2 pose estimation in dynamic environment, an Instance Segmentation method is proposed to remove the moving feature points distributed on the human body and improve the pose accuracy in view of the deception of motion. In this method, ORB feature points are extracted from the input image, and the image is segmented to obtain the position of the pixels in the image. Then the feature points distributed above the human are removed, and the position and attitude are estimated by using the feature points which are relatively stable after the removal. The improved method is used to test on TUM data set. The results show that the improved system can significantly reduce the absolute error and relative drift of pose estimation in dynamic environment, which proves that this method can significantly improve the accuracy of pose estimation in dynamic environment compared with the traditional ORB-SLAM2 system.

 

 

All Content © PaperCept, Inc.

This site is protected by copyright and trademark laws under US and International law.
All rights reserved. © 2002-2022 PaperCept, Inc.
Page generated 2022-01-20  02:19:49 PST  Terms of use