Visual odometry in dynamic environments using light weight semantic segmentation

Visual odometry is the method in which a robot tracks its position and orientation using a sequence of images. Feature based visual odometry matches feature between frames and estimates the pose of the robot according to the matched features. These methods typically assume a static environment and r...

Full description

Saved in:
Bibliographic Details
Main Authors: Tan Ai, Richard Josiah C., Ligutan, Dino Dominic F., Brillantes, Allysa Kate M., Espanola, Jason L., Dadios, Elmer P.
Format: text
Published: Animo Repository 2019
Subjects:
Online Access:https://animorepository.dlsu.edu.ph/faculty_research/405
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: De La Salle University
id oai:animorepository.dlsu.edu.ph:faculty_research-1404
record_format eprints
spelling oai:animorepository.dlsu.edu.ph:faculty_research-14042021-12-10T03:15:56Z Visual odometry in dynamic environments using light weight semantic segmentation Tan Ai, Richard Josiah C. Ligutan, Dino Dominic F. Brillantes, Allysa Kate M. Espanola, Jason L. Dadios, Elmer P. Visual odometry is the method in which a robot tracks its position and orientation using a sequence of images. Feature based visual odometry matches feature between frames and estimates the pose of the robot according to the matched features. These methods typically assume a static environment and relies on statistical methods such as RANSAC to remove outliers such as moving objects. But in highly dynamic environment where majority of the scene is composed of moving objects these methods fail. This paper proposes to use the feature based visual odometry part of ORB-SLAM2 RGB-D and improve it using DeepLabv3-MobileNetV2 semantic segmentation. The semantic segmentation algorithm is used to segment the image, then extracted feature points that are on pixels of dynamic objects (people) are not tracked. The method is tested on TUM-RGBD dataset. Evaluation shows that the proposed algorithm performs significantly better in dynamic scenes compared to the base algorithm, with reduction in Absolute Trajectory Error (ATE) greater than 92.90% compared to the base algorithm in fr3w-xyz, fr3w-rpy and fr3-half sequences. Additionally, when comparing the algorithm that used DeepLabv3-MobileNetV2 to the computationally intensive DeepLabv3-Xception65, the largest increase in ATE was 27%, while the computation time is 3 times faster. © 2019 IEEE. 2019-11-01T07:00:00Z text text/html https://animorepository.dlsu.edu.ph/faculty_research/405 Faculty Research Work Animo Repository Robot vision Computer vision Image segmentation Manufacturing Mechanical Engineering Robotics
institution De La Salle University
building De La Salle University Library
continent Asia
country Philippines
Philippines
content_provider De La Salle University Library
collection DLSU Institutional Repository
topic Robot vision
Computer vision
Image segmentation
Manufacturing
Mechanical Engineering
Robotics
spellingShingle Robot vision
Computer vision
Image segmentation
Manufacturing
Mechanical Engineering
Robotics
Tan Ai, Richard Josiah C.
Ligutan, Dino Dominic F.
Brillantes, Allysa Kate M.
Espanola, Jason L.
Dadios, Elmer P.
Visual odometry in dynamic environments using light weight semantic segmentation
description Visual odometry is the method in which a robot tracks its position and orientation using a sequence of images. Feature based visual odometry matches feature between frames and estimates the pose of the robot according to the matched features. These methods typically assume a static environment and relies on statistical methods such as RANSAC to remove outliers such as moving objects. But in highly dynamic environment where majority of the scene is composed of moving objects these methods fail. This paper proposes to use the feature based visual odometry part of ORB-SLAM2 RGB-D and improve it using DeepLabv3-MobileNetV2 semantic segmentation. The semantic segmentation algorithm is used to segment the image, then extracted feature points that are on pixels of dynamic objects (people) are not tracked. The method is tested on TUM-RGBD dataset. Evaluation shows that the proposed algorithm performs significantly better in dynamic scenes compared to the base algorithm, with reduction in Absolute Trajectory Error (ATE) greater than 92.90% compared to the base algorithm in fr3w-xyz, fr3w-rpy and fr3-half sequences. Additionally, when comparing the algorithm that used DeepLabv3-MobileNetV2 to the computationally intensive DeepLabv3-Xception65, the largest increase in ATE was 27%, while the computation time is 3 times faster. © 2019 IEEE.
format text
author Tan Ai, Richard Josiah C.
Ligutan, Dino Dominic F.
Brillantes, Allysa Kate M.
Espanola, Jason L.
Dadios, Elmer P.
author_facet Tan Ai, Richard Josiah C.
Ligutan, Dino Dominic F.
Brillantes, Allysa Kate M.
Espanola, Jason L.
Dadios, Elmer P.
author_sort Tan Ai, Richard Josiah C.
title Visual odometry in dynamic environments using light weight semantic segmentation
title_short Visual odometry in dynamic environments using light weight semantic segmentation
title_full Visual odometry in dynamic environments using light weight semantic segmentation
title_fullStr Visual odometry in dynamic environments using light weight semantic segmentation
title_full_unstemmed Visual odometry in dynamic environments using light weight semantic segmentation
title_sort visual odometry in dynamic environments using light weight semantic segmentation
publisher Animo Repository
publishDate 2019
url https://animorepository.dlsu.edu.ph/faculty_research/405
_version_ 1719000539668152320