Detection of Round Ban Traffic Signs in Natural Environment

In this paper, a new fast, stable automatic detection method for round ban traffic signs is proposed. Color segmentation based on RGB color space is conducted on sign images. Given the poor adaptability of segmenting by calculating direct difference between channels and setting fixed thresholds, a novel self-adaptive method is proposed that calculates the relative channel difference and fitting threshold curve based on selecting a basic channel. Meanwhile, a gradient filter method separates the signs from its background with the same red color. Edges are extracted and a method of error estimation after least square ellipse fitting screens out the sign edges. Experimental results show that the method presented in this paper can automatically detect round ban traffic signs from sign images of deferent brightness in natural environments, with good prospects in Intelligent transportation as it is stable and applicable in real-time.