基于车道线识别和多特征的前车检测算法

前车检测是安全辅助驾驶系统的主要研究领域,实时鲁棒的检测方法能够使智能车实现有效的防偏防撞预警和控制。提出了一种基于车道线识别和多特征的前车检测方法。首先基于车道线识别方法将感兴趣区域定位于两条车道线之间,然后依据车底阴影特征自适应确定中远距离车辆假设区域,最后利用纹理特征、垂直梯度投影均值和边缘对称性特征三层约束来验证车辆区域。实验结果表明:该方法能够实时准确地检测出不同光照条件下本车车道前方中远距离的车辆。

Preceding Vehicle Detection Algorithm Based on Lane Recognition and Multi-characteristics

Preceding Vehicle Detection is a major research area of Safety Assistant Driving System, real-time and robust detection method can make intelligent vehicle implement the effective early-warning and control of the deviation prevention and anti-collision conditions. This paper proposed a preceding vehicle detection method based on multi-characteristics and lane recognition. Firstly, the region of interest(ROI) was restricted to left and right lane boundaries by using lane detection method. According to the feature of the shadow underneath a vehicle, a self-adjusting bounding box was then used to extract likely mid-range and distance vehicle region. Texture, vertical gradient projection mean and edge symmetry were used to verify the hypothesis region at last. Experiments illustrate the real-time and accurate performance of the detection method for mid-range and distant vehicles in front of self-lane under different illumination condition.