【作者】 吴沫；

【导师】 常文森； 安向京；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2005， 硕士

【摘要】 论文研究了基于计算机视觉的车道跑偏告警系统，提出了一种新的车道跑偏检测方法，并在红旗自主驾驶系统中进行了实验验证。 论文对驾驶员—汽车—道路闭环系统进行了研究，分析了驾驶员的行为如何影响汽车的运动，进而影响车路关系。由这一分析得出通过监测车路关系的变化可以反推驾驶员状态变化的结论，并通过仿真验证了这一结论。 论文系统的介绍了车道跑偏告警系统，包括其性能要求、系统组成模块以及系统评价标准。论文还分析和研究了应用广泛的TLC方法，仿真了由驾驶员状态改变对TLC曲线产生的影响。 论文提出了用于车道跑偏检测的车道线夹角法，不同于TLC方法的是，该方法能够完全基于前视摄像机获得的具有强烈透视效果的图像工作，论文通过数学推导，给出了车道线夹角的表达式及其与普遍使用的TLC方法的关系，说明了车道线夹角法在车道跑偏判定上的可行性。 论文以红旗车自主驾驶系统为实验环境验证了车道线夹角法，给出了包括直道、弯道以及驾驶员不同驾驶习惯下的实验结果及结论。更多还原

【Abstract】 This dissertation has researched on vision-based lane departure warning system. A new lane departure detection approach is proposed, and proved in the HONGQI prototype.A driver-vehicle-road closed loop system is researched. The simulation turned out that the driver’s state can be deduced reversely from the road-vehicle relationship.Lane departure warning system is introduced systematically in the dissertation, including performance demand, system component module, and system evaluate standard. A widely used lane departure detection approach TLC is also analyzed. How the TLC curve varies according to different driver states is simulated.A new criterion, angle between lane markings (ALM), is proposed to detect lane departure. The ALM method can work based on the perspective image. The expression of ALM and its relation to TLC is also presented to prove the feasibility of ALM method.The ALM method has been verified in HONGQI Autonomous Land Vehicle. Results under different conditions are introduced, including straight road, curved road, and different driving customs.更多还原