交通信号灯的检测与识别

Detection and Identification of Traffic Light

交通信号灯的检测与识别系统是无人驾驶与辅助驾驶必不可少的一部分，同时还可以辅助色觉障碍人群与疲劳司机进行安全驾驶，因此本文提出一种能检测和识别交通信号灯状态的新方法。首先在HSVg彩空间对交通信号灯进行颜色分割，利用交通信号灯被黑色矩形框包围的特征进行形状分割，将颜色分割与形状分割相结合确认候选区域；使用颜色直方图统计候选区域的H分量，识别交通信号灯的颜色；使用模板匹配的方法，识别箭头型交通信号灯的方向。实验表明，该方法的检测与识别率达到97％以上。

Because of the traffic signal detection and recogmdon system is an essential part of unmanned systems and auxifiary driving system, at the same time it can also auxiliary dys- chromatopsia crowd and fatigue driver safe driving, so this paper puts forward a way to idend the state of the traffic lights.Segmenting the traffic light in HSV color space.Using the typical characters of the traffic light surrounded by a black rectangle to shape the traffic fight split.According to the shape of segmentation candidate areas to confirm the position of color segmenta- tion.The number of H of labeled region images is counted by color histogram, as basis to judge the traffic fight color.Using template matching, confirming the direction of the arrow traffic lights.Experimental results indicate that the overall reeogrufion rates of the proposed method are over 97%.