道路交通标志识别的研究现状及展望

十几年来，道路交通标志识别的研究工作已经取得了一定成果，但还存在一些不足之处，主要有：识别对象单一，样本数少；处理方法比较单一，智能方法少；偏重于理论的多，面向应用的少；大多数实验对象都是标准图，针对实景图的少；以灰度图为研究对象的多，针对彩色图的少。机器识别面临的主要难点是：道路交通标志的背景相当复杂,颜色失真极为严重并存在不同程度的几何失真；彩色图像处理的理论和技术尚不成熟。“简化复杂问题、改进传统方法、基于颜色信息、采用智能方法”将是今后的一个重要发展方向。

A Survey of the Research on Traffic Sign Recognition

Despite the achievements made in the last decade, the following problems facing the researchers in the field of TSR remain unsolved. First, the objects for recognition research are too few to have generality, training examples remain scant. Second, alternative processing approaches or intelligent approaches has not attracted due attention. Third, application research cannot receive as much attention as that to theories. Fourth, standard or gray images, instead of real or color ones, are mostly studied in experiments. Now the main difficulties in TSR are as follows: first, the traffic sign's background is changeable and not controllable, so color distortion is serious and algebraic distortion exists to various degrees; second, the theory and technology in color image processing need improvement. Adopting intellectual methods and using color information are supposed to be the trends of TSR.