2014
Floros, Georgios; Kyritsis, Konstantinos; Potamianos, Gerasimos
Database and baseline system for detecting degraded traffic signs in urban environments Inproceedings
In: 2014 5th European Workshop on Visual Information Processing (EUVIP), pp. 1–5, IEEE 2014.
Abstract | Links | BibTeX | Tags: Color, Databases, Image color analysis, Roads, Robustness, Shape, Vehicles
@inproceedings{floros2014database,
title = {Database and baseline system for detecting degraded traffic signs in urban environments},
author = {Georgios Floros and Konstantinos Kyritsis and Gerasimos Potamianos},
doi = {10.1109/EUVIP.2014.7018395},
year = {2014},
date = {2014-01-01},
booktitle = {2014 5th European Workshop on Visual Information Processing (EUVIP)},
pages = {1--5},
organization = {IEEE},
abstract = {We present a small database of “noisy” traffic signs in cluttered urban environments that exhibit various forms of degradation, including vandalism and fading (discoloration). The database contains five types of international traffic signs that allow differentiation by means of color and shape, and it has been collected in two cities in Greece. We further present a baseline system for detecting and recognizing signs in this database, primarily employing color segmentation in the RGB color space, shape detection, and a number of problem specific heuristics. Our approach proves quite robust to the degraded traffic signs of our collected database, achieving an F-score of 0.91.},
keywords = {Color, Databases, Image color analysis, Roads, Robustness, Shape, Vehicles},
pubstate = {published},
tppubtype = {inproceedings}
}
We present a small database of “noisy” traffic signs in cluttered urban environments that exhibit various forms of degradation, including vandalism and fading (discoloration). The database contains five types of international traffic signs that allow differentiation by means of color and shape, and it has been collected in two cities in Greece. We further present a baseline system for detecting and recognizing signs in this database, primarily employing color segmentation in the RGB color space, shape detection, and a number of problem specific heuristics. Our approach proves quite robust to the degraded traffic signs of our collected database, achieving an F-score of 0.91.