A driver safety support system which detects traffic lights

Abstract

The number of traffic accidents can be decreased through driver safety support systems (DSSS). In this study, a driver safety support system is proposed in which the driver is warned by finding traffic lights. The proposed DSSS works on only visual information and detects traffic lights. The system primarily transforms the received images into gray scale images and subject them to multi-level thresholding with Otsu criteria. The regions of interest which can be traffic lights are found for the thresholded images by using connected component analysis and blob analysis, respectively. Feature vectors including the color information are extracted from the founded regions. Finally, it is decided if the regions of interest are traffic lights by using support vector machines (SVM). The performance of the proposed DSSS is examined on