Route guidance sign identification using 2-D structural description

We propose a new method to identify route guidance sign (RGS) by comparing the structure data obtained from the RGS database. We describe the structure data extracted from the road scene images taken by an on-board camera. We model RGS as a two level recursive rectangular model. The shape of RGS is modeled as a rectangle, and the characters are represented by circumscribed rectangles. The 3+1 RGS detection modules were used for edge segment extraction to separate the scene from sky region by intensity, for characteristic color extraction in hue saturation value color space, and for template matching using the rectangle of RGS. In each RGS, there are characters including Chinese characters, Roman letters and arrows. We separate the character candidate regions from the sign area using the p-tile method, and represent the character candidate by circumscribed rectangle. The arrows are recognized by thinning and template matching. We applied the proposed method to real road scene images. Experimental results showed the effectiveness of this method.