基于视觉注意机制的交通标志检测

摘要

Two models of visual attention, bottom-up and top-down, which are consistent with human visual perception are introduced to traffic sign detection region. In this paper, it puts forward a way of traffic sign detection based on the visual saliency map. The features of color, shape, intensity and orientation are obtained to make saliency map. It finds the Region of Interest(ROI) by using of the Winners Take All(WTA) network. Experimental results show this method can accurately position the traffic sign in the cases of complex background images.%根据人类视觉感知理论,将自底向上和自顶向下的注意机制相结合,融入到交通标志检测中,提出一种基于视觉注意机制的交通标志检测方法.根据2种注意模型提取颜色、形状、亮度等多种特征,生成显著图,利用WTA网络找到感兴趣区域,即交通标志区域.实验结果表明,该方法能在复杂背景图像中准确定位交通标志.