Item


Feature extraction for underwater visual SLAM

Detecting and selecting proper landmarks is a keyissue to solve Simultaneous Localization and Mapping (SLAM).In this work, we present a novel approach to perform thislandmark detection. Our approach is based on using threesources of information: 1) three-dimensional topological informationfrom SLAM; 2) context information to characterize regionsof interest (RoI); and 3) features extracted from these RoIs.Topological information is taken from the SLAM algorithm,i.e. the three-dimensional approximate position of the landmarkwith a certain level of uncertainty. Contextual information isobtained by segmenting the image into background and RoIs.Features extracted from points of interest are then computed byusing common feature extractors such as SIFT and SURF. Thisinformation is used to associate new observations with knownlandmarks obtained from previous observations. The proposedapproach is tested under a real unstructured underwater environmentusing the SPARUS AUV. Results demonstrate the validityof our approach, improving map consistency

This work was partially funded through the Spanish Ministry of Education and Science (MCINN) under grant CTM2010-15216 and the EU under grant FP7-ICT-2009-248497

Institute of Electrical and Electronics Engineers (IEEE)

Author: Aulinas Masó, Josep M.
Carreras Pérez, Marc
Lladó Bardera, Xavier
Salvi, Joaquim
García Campos, Rafael
Prados Gutiérrez, Ricard
Abstract: Detecting and selecting proper landmarks is a keyissue to solve Simultaneous Localization and Mapping (SLAM).In this work, we present a novel approach to perform thislandmark detection. Our approach is based on using threesources of information: 1) three-dimensional topological informationfrom SLAM; 2) context information to characterize regionsof interest (RoI); and 3) features extracted from these RoIs.Topological information is taken from the SLAM algorithm,i.e. the three-dimensional approximate position of the landmarkwith a certain level of uncertainty. Contextual information isobtained by segmenting the image into background and RoIs.Features extracted from points of interest are then computed byusing common feature extractors such as SIFT and SURF. Thisinformation is used to associate new observations with knownlandmarks obtained from previous observations. The proposedapproach is tested under a real unstructured underwater environmentusing the SPARUS AUV. Results demonstrate the validityof our approach, improving map consistency
This work was partially funded through the Spanish Ministry of Education and Science (MCINN) under grant CTM2010-15216 and the EU under grant FP7-ICT-2009-248497
Document access: http://hdl.handle.net/2072/281190
Language: eng
Publisher: Institute of Electrical and Electronics Engineers (IEEE)
Rights: Tots els drets reservats
Subject: Robots mòbils
Mobile robots
Vehicles submergibles
Submersibles
Imatges -- Segmentació
Image processing
Title: Feature extraction for underwater visual SLAM
Type: info:eu-repo/semantics/article
Repository: Recercat

Subjects

Authors