A visual attention based method for detecting traffic signs of interest

As an important component of the driver assistance system or autonomous vehicle, traffic sign detection can provide drivers or vehicles with safety and alert information about the road. Most existing methods for traffic sign detection only focus on one or several categories of signs while there are various signs in the real world. This paper proposes a biologically-inspired method for detecting almost all categories of traffic signs of interest. Based on the fact that traffic signs are designed to be salient such that they can stand out from its surroundings, this proposed method employs the bottom-up attention mechanism to select the salient objects in the image and the attentional selection is biases based on the top-down attention mechanism so as to filter out non-traffic-sign salient objects. Experimental results have shown that the proposed method is valid for detecting various types of traffic signs.