Road Sign Segmentation based on Colour Spaces: A Comparative Study

U. Zakir, A.N.J. Leonce, and E.A. Edirisinghe (UK)


Colour Segmentation, Road Signs, HSV comparison.


Significant amount of research has already being carried out towards providing practical solutions to automatic road sign detection. Due to the prominence played by colour, an obvious first step of an automatic road sign detection algorithm is colour based segmentation of regions of interest suspected to be road signs. A number of colour spaces have been utilized for this purpose. However due to the variations of colour as a result of illumination changes, bad weather conditions and fading due to exposure to sun-light, makes the threshold selection problem in any colour space a significant challenge. Our literature review revealed that no previous work has been carried out to extensively investigate the robustness of threshold selection, when using different colour spaces in road sign detection. Rather the threshold values have been merely stated and used. In this paper we investigate the use of six colour spaces, specifically recommending the threshold ranges that can be used to robustly segment road signs under different lighting and weather conditions. This work is expected to provide valuable data for further research into robust, automatic road sign detection.

Important Links:

Go Back