【作者】 陈军；

【导师】 王太勇；

【作者基本信息】 天津大学 ， 机械制造及其自动化， 2010， 博士

【摘要】 随着中国汽车工业的快速发展,汽车保有量迅速增加,由此带来的交通事故,特别是恶性交通事故发生率居高不下,交通安全问题日益突出。汽车安全技术研究非常热门,它是车辆工程领域的研究前沿,体现了车辆工程、人工智能、自动控制、计算机等多学科领域理论和技术的交叉与综合。本文从研究车辆主动安全技术出发,在车辆的安全预警领域进行了积极有效的探索。对高速公路车道线识别与跟踪算法、相机标定、车道偏离报警算法进行了系统、深入的研究。为我国汽车安全辅助驾驶系统的产品化提供有力的理论和技术支撑。车道线识别与跟踪是实现车道偏离报警的前提和基础,本文提出一种启发式搜索的车道线识别创新算法,根据梯形匹配模型、车道线灰度变化特征和实际车道宽度约束,确定搜索的起始点,从起始点根据度量代价准则函数搜索车道边界点。采用直线道路模型结合Hough变换来拟合车道边界,通过动态目标搜索区域对车道线进行快速、准确的跟踪。汽车在行驶时会因路面不平产生剧烈颠簸,改变相机与路面的位置关系,对标定结果产生较大的误差。本文根据车辆行驶颠簸的特点以及道路车道线的固有特征,提出了一种车载相机动态误差补偿算法,对车辆颠簸时产生的标定误差进行动态补偿,有效地提高了相机的动态标定精度。分析了空间和时间偏离报警算法的各自特点,充分考虑车辆的运动趋势,提出了基于时间和空间信息融合的偏离报警算法,有效地降低了虚警率,能使系统更好地被驾驶员接受。根据前述研究结果,设计基于BF561的DSP(Digital Singnal Processor)硬件平台、车载相机等组成的单目视觉系统,并进行了数万公里的试验验证,达到产品样机的功能,为我国车道偏离报警系统的广泛应用打下坚实基础。更多还原

【Abstract】 With the rapid develop of vehicle industry in China, the vehicle population increase rapidly. Safety problems have become increasingly prominent for the high incidence of traffic accidents, particularly the serious accidents. Automotive safety technology emerged. It is the forefront of research in the field of vehicle engineering and refers to the vehicle engineering, artificial intelligence, automation, computers and other comprehensive technical fields theory.This article embarks from the vehicle active safety, has carried on some positive beneficial explorations in the vehicles safety warning area of technology. Studying the highway road mark identification and tracking algorithm, camera calibration and departure warning algorithm deeply and systematically. Providing technical and theoretical support for the LDWS (Lane Departure Warning System) becoming a product in China.The identification and tracking of road mark is the precondition and base for LDWS. A innovative road mark identification algorithm based on Heuristic Search is developed in the paper. The starting point is positioned according to the trapezia matching model, gray change property of lane markings and the restraint of real lane width. Then we search the lane edge point from the start point with the tolerance cost rule function. The lane boundary is matched by Hough transform with linear model. Tracking the lane boundary in OSA(Object Search Area) rapidly and accurately.The ubiety between camera and road surface will be changing caused by jolt of vehicle running. It will influence the calibration result badly. According with the inherent characteristic of the road and vehicle jolt feature, a dynamic compensation method for the camera external parameter is proposed in this paper. It can compensate the calibration error caused by jolt in dynamic and improve the calibration precision of camera.After analyzing the each characteristic of departure warning algorithm for time and space, and considering the running trends of vehicle, propose a new departure warning algorithm base on fusion of time and space information. It can lower the false alarming rate and let the LDWS approved by driver more easilyAccording the above research result, design the signal vision system composed of DSP (Digital Singnal Processor) hardware platform of BF561 and camera in-vehicle. It’s function can meet the production prototypes requirement and verified by driving experiment of ten thousands kilometers. It will establish the strong basement for popularizing the LDWS in China.更多还原