夜间复杂交通场景中的车辆检测和跟踪

为提高夜间车辆视频检测和跟踪的准确率，提出一种夜间车辆检测和跟踪算 法.本算法通过亮斑分割和连通组件匹配来检测和定位车辆前灯，并利用区域跟踪算法对 前灯进行跟踪以提高检测准确率.考虑到夜间行车时车辆前灯的显著特征，通过改进Otsu 方法以自适应地分割明亮区域，并根据前灯的几何形状、尺寸及位置信息滤除非车灯部 分的车辆信息.然后利用前灯的对称性进行前灯的配对和归类；最后采用区域跟踪算法对 前灯进行定位和跟踪.实验结果表明，本算法车辆检测平均准确率大于97%，处理速度比 已有方法提高15.8%以上.

Vehicles Detection and Tracking Algorithm for Complex Nighttime Traffic Scene

Vehicles detection and tracking algorithm for nighttime traffic surveillance is proposed in order to further improve the accuracy of video vehicle detection and tracking at night. Light spots segmentation and connect- component matching techniques are used to detect and locate headlights of vehicles and employs region tracking-based method to track headlights. Headlights, which are the only salient features of vehicles in nighttime, are segmented by improved Otsu method, and non-vehicle illumination sources are filtered out according to the geometrical shape, size and location of headlights. Then, headlights are paired and classified based on the geometrical symmetry of headlights. Finally, a region-based tracking algorithm is employed to locate and track headlights. The results prove that the average accuracy rate of the algorithm is more than 97%, and the processing speed is 15.8% higher than the existing