通信论文基于机器视觉的道路检测跟踪系统研究.doc

基于机器视觉的道路检测跟踪系统研究 基于机器视觉的道路检测跟踪系统研究 research of road detection and tracking system based on machine vision 【中文摘要】 基于机器视觉的智能导航系统是智能交通系统的组成部分,是智能车辆导航系统的一项关键技术,而道路图像检测跟踪系统又是智能导航系统中的重点部分,本文针对智能车辆视觉导航系统研究了基于机器视觉的图像处理和分析的道路检测和跟踪算法。论文介绍了智能交通系统的研究背景、研究意义、当今国内外的发展现状、发展趋势和难点;介绍了实现道路检测算法所采用的计算机视觉理论和方法。本文利用视觉传感器(摄像机)采集道路图像,主要利用virtual c+对算法进行仿真。在图像滤波和增强处理方面,经过试验效果对比,特别是通过对比在对图像加入噪声模拟气候条件不好的情况下,发现采用中值滤波效果好、速度快,滤波后边缘也比较清晰,利于以后的继续处理。在图像边缘检测中,通过仿真实验比较,本文利用sobel算子,它的优点是方法简单、计算量小、处理速度快,并且所得的边缘光滑、连续、误差率低,而且sobel算子对噪声条件下的图像检测效果比较令人满意,受到噪声影响小,检测结果边缘比较连续,比较适用于道路图像检测跟踪系统。在本文中,为了提高系统的实时性和鲁棒性,需要对图像进行二值化,选择利用对最佳阈值分割方法进行改进,采用迭代方法,提高了算法的抗噪声能力。为了更好的提取道路图像的车道线,提出了利用基于集合思想的数学形态学,形态学运算保证了道路区域的封闭性,同时对噪声有很好的滤波作用。对于道路图像的车道线,本文采用直线道路模型等道路约束条件,利用hough变换进行直线特征提取,hough变换鲁棒性好,抗噪音性能强,且能连接共线短直线。通过直线参数可以求出车辆相对车道线的位置偏移和角度偏移,接着论文提出利用卡尔曼滤波对车道线跟踪。论文还提出了针对不同路况同时利用gps、红外等异类多传感器进行信息融合,从而达到利用多个传感器联合导航的优势,提高导航系统的有效性。利用采集的实际道路图像对本文提出的算法进行仿真实验,仿真结果表明道路检测和跟踪算法是可行的和有效的。【英文摘要】 intelligent navigation system based on machine vision is a part of the intelligent transportation systems. it is the key technique of intelligent vehicle guidance. the road image tracking system plays an important role in the intelligent navigation system. this paper researches the algorithm of lane detection and track for the visual intelligent vehicle navigation system based image processing and analysis.in this paper, research background, meaning, current status of internal and world, trend and difficulties of the intelligent transportation systems is firstly introduced.then machine vision theory and method for the realization of the road detection are introduced. in this paper, the visual sensor (camera) is used to collect the road images, and virtual c + + is used for the simulation algorithm.in image filtering and enhancement, comparing with the result of the trials, especially comparing the images after adding noise to simulate the real climatic condition, it can be found that the effect of median filtering is fast, good to use, and the edge is clearer then others. it is in favor of the future to deal with. in image edge, by simulate experiments, the paper uses sobel edge detector. it is simple. its operation speed is quick and processes lubricity and series. and it affects by the noise smaller. it is suitable for detection of the road image tracking system. for the real time and robustness of system, the road image must be binary. by the improvement of optical threshold segment, iteration mean is used for the advancement of resist noise. for the distilling road line better, mathematical morphology based on assemble is used for the sealing and filtering of road. abort the line images, as the restriction condition of the line and so on. hough transform is used of distilling the line. the real time and robustness of hough transform is very good. and it links the short ling of one line. by the parameter of road line, the deviation of site and angle are known between the road line and vehicle. and kalman filter is used for tracking the road line.in this paper, it also proposes that using gps, infrared and other heterogeneous multi-sensor information fusion at the same time for different road. the advantage of using different kinds of multiple sensors is that can improve the effectiveness of navigation systems.the algorithm proposed by the paper are used to process actual lane image, simulated results show that the algorithm is right and effective. there are 48 figures and 34 references in this article.【中文关键词】 机器视觉; 图像增强; 边缘检测; hough变换; 卡尔曼滤波 【英文关键词】 machine vision; image enhancement; edge detection; hough transform; kalman filter 【毕业论文目录】 基于机器视觉的道路检测跟踪系统研究致谢 5-6 中文摘要 6-7 abstract 7-8 1 综述 11-18 1.1 课题背景 11-12 1.2 国内外研究现状 12-14 1.3 基于机器视觉的道路检测及研究意义 14-16 1.4 本文所做的工作 16-17 1.5 文章结构 17-18 2 机器视觉检测系统 18-24 2.1 引言 18 2.2 机器视觉及其发展 18-21 2.2.1 机器视觉发展动力 19 2.2.2 机器视觉发展历程与应用趋势 19-20 2.2.3 机器视觉技术环节 20-21 2.3 基于机器视觉的道路检测跟踪系统 21-23 2.3.1 基于机器视觉的道路检测方法 21-22 2.3.2 系统成像几何模型 22-23 2.4 本章小结 23-24 3 道路图像滤波和增强 24-36 3.1 引言 24-25 3.2 图像增强方法 25-35 3.2.1 点处理 26-29 3.2.2 空间域滤波 29-34 3.2.3 频域滤波 34-35 3.3 本章小结 35-36 4 道路图像边缘检测 36-48 4.1 引言 36-37 4.2 微分运算及梯度锐化 37-38 4.3 经典边缘检测算子及效果比较 38-42 4.3.1 roberts边缘检测算子 39 4.3.2 sobel边缘检测算子 39-40 4.3.3 prewitt边缘检测算子 40 4.3.4 kirsch边缘检测 40 4.3.5 高斯-拉普拉斯算子 40-42 4.4 经典边缘提取算子提取图像边缘的结果对比分析 42-47 4.5 本章小结 47-48 5 道路图像检测和跟踪算法研究 48-67 5.1 引言 48 5.2 道路检测算法的假设和算法仿真平台 48-50 5.2.1 算法基本假设 48-49 5.2.2 道路检测软、硬件仿真平台 49-50 5.3 基于道路标识线的道路图像的识