摘要
为提高智能车对多车道的实际道路车辆行驶环境的适应性,提出了一种基于三车道模型的车辆检测方法;方法在预处理的基础上利用极角及位置约束的Hough变换得到可能的车道线信息并利用消失点对车道线进行筛选;利用三车道四线模型对车道线进行匹配;对于每条车道,分别利用车辆灰度信息对车道线内车辆进行识别,并利用视频的连贯性对车辆识别结果进行修正并跟踪车辆;该算法通过对车道线的二次筛选,提高了三车道模型的准确率,进一步提高了对于不同车道车辆识别的正确率;实验结果表明,在结构化道路上,对于不同路况,算法均具有较好的实时性和鲁棒性。
In order to enhance the adaptability to Multi-lane driving environment,a method of car detection based on 3-lane model is proposed.Based on image pre-processing,the potential lane markings are detected using Hough Transform with constraint of angle and location,and then the lane markings are sorted using vanish point.After that,the lane markings are matched to a 3-lane model.For each lane,cars are identified using gray scale information,and the cars are corrected and tracked at last according to video streaming.The accuracy of the 3-lane model and the car identification is improved with the method for secondary classification of lane markings.Experiment verifies that the method has a good robustness and stability for various kinds of lanes on structural road.
出处
《计算机测量与控制》
2015年第11期3736-3738,3743,共4页
Computer Measurement &Control
基金
北京市教委科技创新平台项目(JJ 002790200802)
