《大学毕业论文-—基于adaboost和svm的交通标志识别研究与实现》由会员分享，可在线阅读，更多相关《大学毕业论文-—基于adaboost和svm的交通标志识别研究与实现（68页珍藏版）》请在金锄头文库上搜索。
1、摘 要 硕士学位论文基于AdaBoost和SVM的交通标志识别研究与实现VIIResearch and Implementation of Traffic Sign Recognition Based on Adaboost and SVM论文题目：基于AdaBoost和SVM的交通标志识别研究与实现摘 要交通标志的识别是智能交通标志的重要组成部分。它涉及传感器技术、信息技术、自动化技术和计算机等多种技术以及如何识别道路、识别碰撞、识别交通标志等多种欲识别的对象。经过国内外学者的多年研究，交通标志的识别理论和技术体系已经取得了突破性的进展。一般来说，交通标志的图像的采集是实现智能交通的第一步，
2、它对后续的各项操控是否正确有效至关重要，但交通标志全都暴露在特殊的室外环境中，为使驾驶员看清楚各类交通标志，通常交通标志要放在道路旁和管弯处。在这些地方的交通标志常常容易受到强烈光照、灰尘和树木等多方面的影响，所以图像的清晰度较差，从而影响摄像机对交通标志的采集，车内的嵌入式计算机软硬件系统所接收的图像信息也就模糊不清。正因为如此，人们一直都在致力于如何提高交通标志图像识别率的研究。本论文就如何提高交通标志的识别率进行了一些相关的研究，其研究成果虽然距离实用还有相当大的距离，但其研究过程是使自己开阔了眼界，增长了知识，提高了业务水平。本文的主要研究内容由以下三部分组成：第一部分：交通标示识别数
3、据集确定。介绍了两种图像预处理方法，结合试验进行比对分析；综述了三种交通标志检测方法：基于颜色、形状以及综合两种的检测算法；对各种交通标志特征提取方法进行实验对比分析，实验证明三角形标志和圆形标志被识别错误的概率最高；第二部分，在研究了现有交通标志识别方法AdaBoost和SVM的特点后，采用了一种变的AdaBoost技术、综合颜色和形状的交通标志检测方法、子模式组合的特征提取方法，在子模式的基础上，对比了相邻分块、交叠边缘分块和滑动分块方法和基于径向基核函数的支持向量机分类器相结合的识别方法来识别常见的交通标志。第三部分，论文采用MATLAB软件工具对交通标志识别方法和识别过程进行了设计实现
4、，包括系统的运行环境、业务流程、系统识别图像过程、获取特征向量过程，并进行仿真的对比分析，结果表明，通过改变有关参数和融合AdaBoost和SVM的交通标志的识别方法识别效果更好，识别率更高。关 键 词：交通标志识别，分块核函数，SVM，AdaBoost论文类型：应用研究Title： Research and Implementation of Traffic Sign Recognition Based on Adaboost and SVMSpecialty：Computer Science and TechnologyApplicant：Xinjun ChenSupervisor：Pro
5、f.Xianglin MiaoABSTRACTTraffic sign recognition is an important part of intelligent traffic signs. It involves many kinds of the technology such as sensor technology, information technology, automation technology and computer technology, and how to identify road, identification of collision, identif
6、y the object recognition of traffic signs, etc. After years of research of scholars both at home and abroad, and traffic sign recognition theory and technology system has made breakthrough progress. therefore, the image collection is the first step to realize intelligent transportation, it is very i
7、mportant to the follow-up of the manipulation, but the traffic signs are all exposed to special outdoor environment, to make the drivers see all kinds of traffic signs, traffic signs usually should be placed beside the road and pipe bend. Where traffic signs are often vulnerable to affect by the str
8、ong light, dust and various trees, so the sharpness of image is bad, which affect the camera collection, the cars embedded computer software and hardware system of image information is ambiguous. Because of this, people have been trying to research how to improve the traffic sign image recognition.
9、This paper has discussed some related research how to improve the recognition rate of traffic sign, although the research results have a considerable distance from the practical, but its research process broads the horizons, increases of knowledge, improves the level of the business.The research con
10、tents of this paper include:In the first part, ensures the data sets of traffic sign recognition, introduces two methods of image preprocessing, compares with the combination of experiment analysis; Three traffic sign detection methods are reviewed, based on color, shape, and integrated two detectio
11、n algorithm;In the second part, this paper adopt a variable AdaBoost technology , comprehensive test method of colors and shapes of traffic signs , sub-pattern combination method of feature extraction after research the characteristics of SVM and AdaBoost, on the basis of subschema, compare the edge
12、 block and the adjacent block, overlapping sliding block method and based on the radial basis kernel function of support vector machine classifier combination of identification methods to identify common traffic signs.In the third part, the paper uses the MATLAB software tool for traffic sign recogn
13、ition method and recognition process design and implementation, including the system running surroundings, the process of business, system identification, image process, obtain eigenvector, and contrastive analysis of the simulation, Results show that change the parameters of several identification
14、methods and the comprehensive recognition effect is better, higher recognition rate.KEY WORDS：Traffic Sign Recognition, Block Kernel Function, SVM, AdaBoostTYPE OF THESIS：Applied Research 目 录目 录1 绪论11.1研究的背景和意义11.2国内外研究现状21.3论文工作31.4本文的框架和研究内容42 交通标志识别的实验数据集和图像预处理52.1实验数据集52.2交通标志图像预处理72.2.1随机噪声的消除7
15、2.2.2运动模糊的消除82.3实验结果92.4本章小结113 交通标志的检测123.1基于颜色的交通标志检测123.1.1 RGB颜色空间模型123.1.2 HIS颜色空间模型133.2基于形状的交通标志检测143.3综合颜色与形状的算法153.4本章小结164 交通标志的特征提取174.1基于BKFDA的特征提取174.2基于spBKFDA的特征提取184.2.1 spBKFDA原理184.2.2算法改进194.3实验结果和分析224.3.1基于BKFDA特征提取的分类实验224.3.2基于spBKFDA特征分类实验234.4本章小结255 融合ADABOOST和SVM的交通标志识别265.1基于SVM的交通标志识别265.1.1基于SVM的线性可分和线性不可分265.1.2基于SVM的多类分类275.1.3传统SVM存在的不足295.1.4 SVM算法的改进295.2基于径向基核函数的PTSVM的交通标志识别325.3融合AdaBoost和SVM的交通标志识别345.3.1 Boosting算法345.3.2 AdaBoost算法355.3.3变的AdaBoost算法355.4本章小结386 交通标志的识别与运行分析39
