【作者】 钱慧佳；

【导师】 杨明；

【作者基本信息】 上海交通大学 ， 控制理论与控制工程， 2010， 硕士

【摘要】 行人安全作为智能交通系统中的重要一环,越来越受到各个国家的重视,而对他的检测则是确保其安全的首要目标。目前的行人检测算法中,普遍采用单个传感器的方法,比如单目视觉和激光雷达。本论文以智能车为平台,采用摄像头和激光雷达融合的方法实现行人检测,并对多车协作实现行人检测展开了一定的研究。本文一共分为四个部分。第一个部分主要介绍了摄像头传感器与激光雷达传感器之间联合标定的过程以及这两种传感器同步的方法,它是实现兴趣区域提取的关键步骤,通过两种传感器之间的协作降低算法复杂度。第二部分介绍了行人检测的主要算法,通过联合标定确定的候选区域,通过图像特征和基于Hausdorff距离的模板匹配方法定位行人。第三部分提出了一种基于多车协作来实现行人检测的方法,它可以实现车辆在被遮挡情况下仍能检测到行人并进行预警。本文分别采用GPS和道路信标两种方法来实现。第四部分主要介绍了对道路周边环境的辅助检测,包括停车线检测和道路交通标志检测,它对于进一步确保行人安全有一定的实际意义。更多还原

【Abstract】 More and more countries pay attention to pedestrian safety for its importance in intelligent traffic system. Then, pedestrian detection is the main target in order to ensure their safety. The universal algorithm for pedestrian detection is based on single sensor such as monocular or laser scanner. This paper introduces a multi-sensor fusion method to realize pedestrian detection based on camera and laser scanner which are installed in intelligent vehicle. This paper also has a research on multi-car cooperation to realize pedestrian detection.This paper is divided into four sections. The first part introduces a method of the co-calibration and synchronization between a camera and laser scanner. A proper ROI(region of interests) can be located based on this method which can reduce the algorithm complexity. The second part introduces the main algorithm of pedestrian detection. Through the ROI induced from co-calibration, a method for pedestrian detection is introduced based on image characteristics and a template matching method of Hausdorff distance. The third part presents a pedestrian detection way based on the cooperation between car and car. It can realize pedestrian detection when the car is occluded and warn the pedestrian. This paper uses GPS and traffic beacon to carry out this method. The fourth part introduces the additional detection for road environment。It includes stop line and traffic sign detection. It is helpful for pedestrian safety.更多还原