基于自适应伽马变换的交通标志检测方法

Traffic Sign Detection Method Based on Adaptive Gamma Correction

为提高自然场景下交通标志检测精度,提出一种基于自适应伽马变换的交通标志检测方法。方法利用颜色和形状等特征实现交通标志检测。针对传统交通标志检测方法中较少考虑自然场景图像亮度校正问题,首先提出了基于自适应伽马变换的颜色增强方法,然后采用最大稳定极值区域(MSER)算法提取颜色增强图中的交通标志感兴趣区域(ROI),最后基于方向梯度直方图(HOG)特征,利用级联支持向量机(SVM)对交通标志ROI进行分类判决。基于德国交通标志数据库(GTSDB)的检测实验获得了97.28%的检测率及10.35%误检率,结果表明,所提算法可有效提升交通标志检测的精度。

To improve the traffic sign detection accuracy in the natural environment,a traffic sign detection meth-od based on adaptive gamma correction algorithm is proposed in this paper,which detects the traffic signs via the fea-tures of the color and shape.Aiming at the shortcoming that traditional traffic sign detection methods are less consid-ering of luminance correction for natural scene image,we first designed an effective color enhancement method through adaptive gamma correction.Then,the region of interest(ROI)of traffic signs was extracted from color en-hancement images based on the maximally stable extremum regions(MSER)method.Finally,a cascade support vector machine(SVM)classification algorithm via the histogram of oriented gradient(HOG)feature extracted only in the MSER regions was applied to classify the MSER regions.Experimental results demonstrate the accuracy of the proposed approach on traffic sign detection,and demonstrate the effectiveness and robustness of the approach on traf-fic sign detection.Based on the German Traffic Sign Detection Benchmark Database(GTSDB),97.28%detection rate and 10.35%false detection rate are obtained.