Visual Simultaneous Localization and Mapping for a tree climbing robot
"This work addresses the problem of generating a 3D mesh grid model of a tree by a climbing robot for tree inspection. In order to generate a consistent model of the tree while climbing, the robot needs to be able to track its location while generating the model. Hence we explored this problem...
Main Author: | |
---|---|
Other Authors: | |
Format: | Others |
Published: |
Digital WPI
2013
|
Subjects: | |
Online Access: | https://digitalcommons.wpi.edu/etd-theses/1038 https://digitalcommons.wpi.edu/cgi/viewcontent.cgi?article=2037&context=etd-theses |
id |
ndltd-wpi.edu-oai-digitalcommons.wpi.edu-etd-theses-2037 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-wpi.edu-oai-digitalcommons.wpi.edu-etd-theses-20372019-03-22T05:45:46Z Visual Simultaneous Localization and Mapping for a tree climbing robot Wisely Babu, Benzun Pious "This work addresses the problem of generating a 3D mesh grid model of a tree by a climbing robot for tree inspection. In order to generate a consistent model of the tree while climbing, the robot needs to be able to track its location while generating the model. Hence we explored this problem as a subset of Simultaneous Localization and Mapping problem. The monocular camera based Visual Simultaneous Localization and Mapping(VSLAM) algorithm was adopted to map the features on the tree. Multi-scale grid based FAST feature detector combined with Lucas Kande Optical flow was used to extract features from the tree. Inverse depth representation of feature was selected to seamlessly handle newly initialized features. The camera and the feature states along with their co-variances are managed in an Extended Kalman filter. In our VSLAM implementation we have attempted to track a large number of features. From the sparse spatial distribution of features we get using Extended Kalman filter we attempt to generate a 3D mesh grid model with the help of an unordered triangle fitting algorithm. We explored the implementation in C++ using Eigen, OpenCV and Point Cloud Library. A multi-threaded software design of the VSLAM algorithm was implemented. The algorithm was evaluated with image sets from trees susceptible to Asian Long Horn Beetle. " 2013-09-19T07:00:00Z text application/pdf https://digitalcommons.wpi.edu/etd-theses/1038 https://digitalcommons.wpi.edu/cgi/viewcontent.cgi?article=2037&context=etd-theses Masters Theses (All Theses, All Years) Digital WPI Taskin Padir, Committee Member David Cyganski, Committee Member Michael A. Gennert, Advisor FAST grid Extended Kalman Filter scansorial robots VSLAM |
collection |
NDLTD |
format |
Others
|
sources |
NDLTD |
topic |
FAST grid Extended Kalman Filter scansorial robots VSLAM |
spellingShingle |
FAST grid Extended Kalman Filter scansorial robots VSLAM Wisely Babu, Benzun Pious Visual Simultaneous Localization and Mapping for a tree climbing robot |
description |
"This work addresses the problem of generating a 3D mesh grid model of a tree by a climbing robot for tree inspection. In order to generate a consistent model of the tree while climbing, the robot needs to be able to track its location while generating the model. Hence we explored this problem as a subset of Simultaneous Localization and Mapping problem. The monocular camera based Visual Simultaneous Localization and Mapping(VSLAM) algorithm was adopted to map the features on the tree. Multi-scale grid based FAST feature detector combined with Lucas Kande Optical flow was used to extract features from the tree. Inverse depth representation of feature was selected to seamlessly handle newly initialized features. The camera and the feature states along with their co-variances are managed in an Extended Kalman filter. In our VSLAM implementation we have attempted to track a large number of features. From the sparse spatial distribution of features we get using Extended Kalman filter we attempt to generate a 3D mesh grid model with the help of an unordered triangle fitting algorithm. We explored the implementation in C++ using Eigen, OpenCV and Point Cloud Library. A multi-threaded software design of the VSLAM algorithm was implemented. The algorithm was evaluated with image sets from trees susceptible to Asian Long Horn Beetle. " |
author2 |
Taskin Padir, Committee Member |
author_facet |
Taskin Padir, Committee Member Wisely Babu, Benzun Pious |
author |
Wisely Babu, Benzun Pious |
author_sort |
Wisely Babu, Benzun Pious |
title |
Visual Simultaneous Localization and Mapping for a tree climbing robot |
title_short |
Visual Simultaneous Localization and Mapping for a tree climbing robot |
title_full |
Visual Simultaneous Localization and Mapping for a tree climbing robot |
title_fullStr |
Visual Simultaneous Localization and Mapping for a tree climbing robot |
title_full_unstemmed |
Visual Simultaneous Localization and Mapping for a tree climbing robot |
title_sort |
visual simultaneous localization and mapping for a tree climbing robot |
publisher |
Digital WPI |
publishDate |
2013 |
url |
https://digitalcommons.wpi.edu/etd-theses/1038 https://digitalcommons.wpi.edu/cgi/viewcontent.cgi?article=2037&context=etd-theses |
work_keys_str_mv |
AT wiselybabubenzunpious visualsimultaneouslocalizationandmappingforatreeclimbingrobot |
_version_ |
1719005880334155776 |