基于团块几何和位置特征的夜间前方车辆检测方法

Preceding vehicle detecting method for night based on geometry and position features of blobs

夜间前方车辆检测是智能车辆安全辅助驾驶的关键技术.提出了一种基于单目视觉的夜间前方车辆检测方法.首先对视频图像进行预处理得到二值图像,然后提取二值图像中的团块及团块的几何和位置特征,最后根据车辆尾灯对及车辆尾部的大面积反光区域的几何和位置特征,建立生成假设车辆的判断机制,用以检测出前方车辆.使用实际采集的道路视频进行测试,结果表明该方法能够在夜间有效地检测出前方车辆.

The detection of preceding vehicles at night is the key technique for safety assistant system in intelligent vehicle driving. This paper presents a preceding vehicle detection method at night based on monocular vision. The first step is image preprocessing. The video image is processed to obtain the binary image. Then the blobs and the features, geometry and position of blobs are extracted from the binary image. At last, in order to detect the preceding vehicle, a judgment mechanism is established to generate hypothetical vehicles based on geometry and position features of the rear light or the large reflective area of the rear vehicle. The method is tested with practical video taken on road. The results show that this method can effectively detect preceding vehicles at night.