Road Sign Segmentation and Recognition under Bad Illumination Condition using Modified Fuzzy C-means Clustering

Abstract

In this paper, we present a novel approach on road sign segmentation under bad lighting condition employing a modified fuzzy c-means clustering. Most of the cases accidents occur in bad weather or when the road signs cannot be recognized. The proposed system implements a system that segments the road sign using fuzzy c-means clustering. At first, the image is detected in RGB colour space and then converted into HSV colour model. After the conversion we detected the edge by applying Canny edge detector. Thus the proposed method is applied on the image. The method is based on the local similarity measure so that the noise created or the blurred image can easily be detected. But the traditional Fuzzy C-means clustering lack enough robustness to noise and preserving details of the image. Experimental results demonstrate that the system can segment the road signs successfully under bad illumination conditions. This system can be applied to typical type of traffic signs such as triangular, circle etc. The output result of the system is encouraging as the accuracy rate is 99%.