复杂场景下基于颜色对和MSER的交通标志检测方法

Detection method of traffic signs based on color pair and MSER in the complex environment

交通标志识别技术在室外复杂环境中,会有光照变化、标志褪色、标志倾斜和形变等不利因素的影响.目前,主流的检测算法使用颜色、形状或局部稳定特征进行交通标志的检测.但这些算法在复杂场景下检测精度低,鲁棒性差.为了达到较好的检测效果,提出一种基于颜色对和最大稳定极值区域(MSER)的交通标志检测算法.在检测过程中,为了解决不同天气情况下获取的图像亮度不同的问题,使用多组阈值;针对使用传统的颜色阈值算法得到的交通标志候选区域与颜色相似的背景粘连在一起的现象,提出一种分割算法对其进行准确分割.最后使用支持向量机(SVM)和方向梯度直方图(HOG)对得到的目标区域进行分类.尽管是在复杂的场景下,交通标志的检测率和识别率也较高.

Traffic signs recognition technology can be used in driving assistance, automatic drive, and routing maintenance of traffic sign. This technology has a significant research value. However, in the complex outdoor environment, due to the unfavorable factors of variable light condi tion, signs fading, signs inclining and distorting, traffic signs detection and recognition is still a challenging task. Currently,color, shape and local stability are used in the main methods of traffic sign detection. However, these methods have the disadvantages of low precision and less robust ness in the complex environment. To achieve a better detection result, a new method for traffic sign detection is proposed which based on color pair and Maximally Stable Extremal Regions（MSER）. In the procedure of detecting traffic signs, a multi threshold method is utilized for the prob lem of different image brightness that obtained in the different weather condition. And a cutting method is also proposed for the candidate regions which are connected with similar background. At last, SVM and HOG are utilized for the classification of the proposed regions and high detec tion rate and recognition rate are obtained in the complex environment.