Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network

  Advanced Search  

FLIRT: Interest Regions for 2D Range Data with Applications to Robot Navigation

Summary: FLIRT: Interest Regions for 2D Range Data with
Applications to Robot Navigation
Gian Diego Tipaldi, Manuel Braun, and Kai O. Arras
Abstract In this paper, we present FLIRT that stands for Fast Laser Interest Region
Transform, a multi-scale interest region operator for 2D range data. FLIRT com-
bines a detector based on a geodesic curve approximation of the range signal and a
descriptor based on a polar histogram of occupancy probabilities. This combination
was found in a set of comparative benchmark experiments on standard indoor and
outdoor data sets. The analysis yields repeatability and matching performance re-
sults similar to the values found for interest points in the computer vision literature,
encouraging a wide application of FLIRT on 2D range data. We show how FLIRT in
conjunction with RANSAC can be applied for loop closing, global localization, in-
cremental mapping and pose-based SLAM with particularly simple algorithms. The
results demonstrate that FLIRT features have a great potential for robot navigation
in terms of precision-recall performance, efficiency, generality and simplicity.
1 Introduction
The introduction of local image features had a large impact on many computer vi-
sion tasks such as object and scene recognition, motion tracking, stereo correspon-
dence, or visual robot localization and SLAM. The typical strategy is to select lo-
cations of interest in image space and compute a distinctive descriptor over regions


Source: Arras, Kai O. - Institut für Informatik, Albert-Ludwigs-Universität Freiburg


Collections: Computer Technologies and Information Sciences