Abstract
A traffic sign detection system in the vehicle can be of great help
for the driver. The number of accidents can be reduced by 20\% if the
speed limits are followed. A system that warns the driver about speeding could therefore save lives if the driver reduces the speed.
This work focus on the colour classification used in traffic sign
detection methods. Existing methods are compared, and a Genetic
Algorithm is used for optimising parameters used in the existing
colour classification methods.
Cartesian Genetic Programming is used for evolving colour classifiers
for traffic signs, and compared to the existing methods. The evolved
classifier is tested with three different luminance adjustment
algorithms.
The results show that the GA is able to find better parameters than
the reported parameters, and some of the evolved colour classifiers were
better than the existing methods. The CGP architecture did find better
classifiers than the existing. The luminance adjustment algorithms did
not result in better classification results.