道路场景中交通标志的检测方法

为在车辆视觉导航中能快速识别道路场景中的交通标志,通过提取标志的特征颜色与识别标志的特征形状来实现交通标志的检测。将道路场景RGB空间图像转换为HSI空间,利用色调与饱和度融合的办法来提取道路场景中标志的特征颜色区域,二值化后排除噪声并通过投影法得到正确的标志区域,实现交通标志的定位。以红色禁令圆形交通标志为例,提出一种改进的Hough变换方法来识别标志域内的几何特征——圆形标记。通过对不同条件下20幅含禁令标志的道路场景图像的检测试验,发现本方法对交通标志颜色提取定位与几何特征图形识别的正确率都达到100%,且平均检测时间为245ms。分析结果表明:该方法能够快速、准确地确定标志区域,同时对禁令标志几何图形识别具有较强的实时性和鲁棒性。

Detection method of traffic signs in road scenes

In order to quickly identify traffic signs in road scenes with vehicle visual navigation,a detection method was put forward by extracting the color characteristics of traffic signs and recognizing the shape characteristics of targets. The road scene image of RGB space was transformed into that of HSI space,a syncretic method of hue and saturation was used to extract the characteristic color area of sign in road scene,then traffic sign image was binarized,the noise was removed,and binarization target was projected to locate traffic sign. A red circular interdictory traffic sign was taken as example,and an improved Hough transform method was proposed to recognize the geometrics of the sign area. Analysis result shows that after 20 road scene tests,including interdictory traffic signs under different conditions,the detection rates of color area locations and geometric characteristics recognitions for traffic signs reach 100%,and the average detection time is 245 ms,so the method not only locates the sign area quickly and correctly,but also recognizes the geometric graph of interdictory traffic sign with real-time property and robustness. 1 tab,9 figs,16 refs.