基于角点检测估算车辆间即碰时间.PDF

ISSN 1674-8484 汽车安全与节能学报, 2014 年, 第5 卷 第4 期 03/ 15 CN 11-5904/U J Automotive Safety and Energy, 2014, Vol. 5 No. 4 331—335 基于角点检测估算车辆间即碰时间 朱西产，高学敏，许宇能，李  霖 （同济大学 汽车学院，上海201804 ，中国） 摘  要： 车辆间即碰时间（TTC）是驾驶辅助系统控制策略的基础，也是分析驾驶员驾驶行为的主要依 据。为分析行车记录仪视频危险工况中驾驶员紧急制动时的TTC，提出了一种基于单目视觉估算算法。 该算法先检测及跟踪车辆，判断前方有无车辆及其具体方位；再检测并匹配车辆角点，计算不同帧 之间车辆图像的尺寸变化率，进而估算 TTC。用最小二乘法、Kalman 滤波方法处理数据；用多层金字 塔图像来处理图像，实现了大运动的角点匹配。结果表明：该算法丰富了视频危险工况参数的分析方法， 拓展了单目视觉的应用场景，为低成本驾驶辅助系统，提供了新的解决方案。 关键词： 汽车工程；主动安全；危险工况；车辆间即碰时间(TTC)；驾驶辅助系统；角点检测 中图分类号： U 463.3 文献标识码： A DOI: 10.3969/j.issn.1674-8484.2014.04.003 Time to collision between vehicles estimation based on corner detection ZHU Xichan, GAO Xuemin, XU Yuneng, LI Lin (School of Automotive Studies, TongJi University, Shanghai 201804, China) Abstract: Time to Collision (TTC) between vehicles is used to control strategy of driver assistance systems, and to analyze drivers’ behavior. A calculation approach was presented to estimate the TTC of emergency braking in risk scenarios using a monocular camera. The vehicle in front of the host car is detected and tracked, followed with calculating the observed vehicle scale change based on the corner detection and using the scale change to estimate the TTC. Least square method and Kalman filtering were used in processing data, with the image pyramid being applied to realize the corner matching in large frame movements. The results show that the algorithm is a new tool for analyzing the risk scenarios, while expanding monocular camera application and providing a new solution for low-cost driver assistance systems. Keywords: a