Authors:
Pangyu Jeong
and
Sergiu Nedevschi
Affiliation:
Technical University of Cluj-Napoca, Romania
Keyword(s):
Visual Odometry (VO), Ego Motion Model, Ego Motion Control, Localization, Map Building, Temporal Filter.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Computer Vision, Visualization and Computer Graphics
;
Human-Computer Interaction
;
Methodologies and Methods
;
Motion and Tracking
;
Motion, Tracking and Stereo Vision
;
Pattern Recognition
;
Physiological Computing Systems
;
Real-Time Vision
;
Stereo Vision and Structure from Motion
;
Visual Navigation
;
Visually Guided Robotics
Abstract:
This paper presents a robust method for localization and map building in dynamic environment. The proposed localization and map building provide general approaches to use them both in indoor and outdoor environments. The proposed localization is based on the relative global position starting from initial departure position. In order to provide reliable positioning information, Visual Odometry (VO) is used instead of ego robot’s encoder. Unlike general VO based localization, the proposed VO does not use iterative refinement in order to select inliers. The suggested VO uses ego motion model based on the motion control. The rotation and translation values of tracked features are guided by the estimated rotation and translation values obtained by motion control. Namely the estimated motion provides upper and lower limits of motion variation of VO. This estimated boundary of motion variation helps to reject outliers among tracked features. The rejected outliers represent tracked features
of fast/slow moving objects against ego robot movement. The map is built along with ego robot path. In order to get rich 3D points in each frame accumulated dense map based temporal filter method is adapted.
(More)