【作者】 刘波；

【导师】 金施群；

【作者基本信息】 合肥工业大学 ， 精密仪器及机械， 2009， 硕士

【摘要】 随着现代社会交通的发展,随之而来的日渐紧张的交通状况成为最直接的安全隐患,各类交通事故明显增加。车辆防撞预警系统作为智能交通系统的核心部分,正成为当今世界车辆工程领域的研究热点。通过对国内外智能车辆防撞预警系统的研究分析发现,现有的基于视觉检测的系统在阴雨、大雾等较为恶劣的天气情况下检测结果不太满意。因此,本文提出了基于红外视觉检测技术的汽车防撞预警系统设计方案。文中首先对几种红外图像采集系统的性能给予分析,构建了主动式红外图像采集系统。该系统主要由红外LED灯、低照度黑白CCD摄像头、截止型红外滤光片等构成。根据采集的红外道路图像的特点和先验知识,本文提出用等腰梯形来分割图像,并在此基础上研究了红外图像预处理方法,包括:中值滤波、Sobel边缘检测和最大最小方差阈值分割。然后在确定的范围内运用Hough变换检测车道线边缘,进而确定感兴趣区域(AOI)。接着在感兴趣区域内运用灰度变化的原理搜索前方车辆,并确定其在图像中的位置;运用卡尔曼滤波技术实现了对前方车辆的实时跟踪。最后基于投影几何模型,利用道路平坦假设,建立了世界坐标系、摄像机坐标系、图像坐标系及帧存坐标系之间的坐标变换方程,推导出本车与前方车辆距离计算公式并提出安全距离判断准则。经过以上一系列处理,本系统能实时、准确地检测到本车道内的前方车辆并确定其位置,实现车辆防撞预警功能。更多还原

【Abstract】 With the development of traffic in the modern society, increasing busyness of traffic conditions has become the most direct safety problems, and various types of accidents increased. As a hard-core of the Intelligent Transportation System, the in-cars pre-warning system is very hot in the research field of vehicle engineering in today’s world.Through analyzing the domestic and international research of the in-cars pre-warning system, we found that these vision systems doesn’t work well in rain, fog and other adverse weather conditions. So in this paper we designed the in-cars pre-warning system which based on the infrared vision detection.Firstly, active infrared image collection system was designed after analysis the performance of several infrared image collection systems. It mainly makes up of infrared LED light, low-luminance CCD camera and infrared filter. Secondly, based on the characters of the infrared road images and the priori knowledge, isosceles trapezoid was used to segment the image. Then we studied the pre-processing method of infrared images which contains median filtering, Sobel edge detection and the largest minimum variance threshold segmentation. Next, the area of interest (AOI) was determined after detect the edge of lane line by Hough transformation. The system use the principle of gray change to search the front cars in the area of interest, use Kalman filtering to track front cars. Finally, based on the geometric model projection, flat road hypothesis is used to set up the coordinate transformation equations between the world coordinate system, camera coordinate system, image coordinate system and frame buffer coordinates, then derived the formula for measure distance from two cars and the safe distance criterion.After the process above, this system can detect the front cars and site them quickly and accurately, implement the in-cars pre-warning system function.更多还原