基于消失点的鲁棒逆透视变换重庆大学硕士学位论文(学术学位)学生姓名:张代明指导教师:房斌教授专业:计算机应用技术学科门类:工学重庆大学计算机学院二O一五年四月Robust Inverse Perspective Mapping Based on Vanishing PointA Thesis Submitted to Chongqing UniversityIn Partial Fulfillment of the Requirement for theMaster’s Degree of EngineeringByZhang DaimingSupervised by Prof. Fang BinSpecialty:Computer Application TechnologyCollege of Computer Science ofChongqing University, Chongqing, ChinaApril 2015摘要逆透视变换在计算机视觉和道路交通标志检测和识别方面得到了广泛的应用。
逆透视变换是透视变换的逆过程,主要是结合相机的内在和外在参数,将图像从图像坐标系映射到世界坐标系,从而消除透视影响对图像检测和识别任务的干扰和误差。
因此,鲁棒的逆透视变换对于透视影响的消除和获取图像信息的不变量有着非常重要的作用。
本文主要研究了基于消失点的鲁棒逆透视变换及其应用,采用消失点对逆透视变换需要的部分偏转角度参数进行了自动计算,并将逆透视变换应用到导向箭头的检测和识别任务中。
首先,本文介绍了基于消失点的鲁棒逆透视变换的选题背景以及研究意义,总结了现有逆透视变换算法的国内外研究现状以及存在的难点问题。
其次,本文介绍了逆透视变换的基本概念及其数学原理,总结了逆透视变换在道路交通标线检测的广泛应用,并对常见逆透视变换方法进行了分类总结。
从基于点变换的逆透视变换、基于矩阵变换的逆透视变换和基于简化相机模型推导的逆透视变换三个方面介绍了现有逆透视变换的优点和缺点。
其中,本文主要介绍了基于简化相机模型推导的逆透视变换,该类逆透视变换方法计算简单,具有很高的实用价值。
再次,本文提出了一种基于消失点检测的鲁棒逆透视变换方法。
消失点反映了图像的透视结构,可以采用消失点坐标对相机俯仰角和偏航角进行计算,为逆透视变换提供实时偏转角度参数,增强逆透视变换对于上下坡等非常规道路环境的鲁棒性。
同时,改进了一种逆透视变换方法,加上了偏航角矫正,提高了道路平面俯视图的质量。
本文首先检测出消失点坐标,然后根据消失点计算相机俯仰角和偏航角,最后通过逆透视变换得到道路平面的俯视图,消除道路平面的透视形变。
最后,本文将逆透视变换应用到导向箭头的检测和识别当中。
导向箭头作为一种道路交通标线,对车辆的规范行驶有着重要的指示作用。
因此,导向箭头的检测与识别可以为智能交通系统提供重要的导向信息。
本文首先通过逆透视变换消除透视影响从而得到道路平面的俯视图,然后检测出导向箭头候选区域,最后对候选区域采用形状上下文进行形状特征提取,并结合基于蚁群优化的形状匹配算法来识别导向箭头。
关键词:消失点检测,逆透视变换,智能交通,导向箭头检测与识别,形状匹配ABSTRACTInverse perspective mapping (IPM) has been widely used in computer vision and road traffic makings detection and recognition. Inverse perspective mapping is the inverse process of perspective mapping. It maps the image from the image coordinates to the world coordinate by a combination of intrinsic and extrinsic parameters of the camera, and eliminate the perspective effect. Therefore, a robust inverse perspective mapping method plays a very important role in eliminating the perspective influence and obtaining the invariant information of the image. This paper studies the robust inverse perspective mapping methods based on the vanishing point and its application. It calculates the deflection angle of the inverse perspective mapping parameters based on the vanishing point automatically, and then uses the inverse perspective mapping to detect and recognize the arrow road markings.Firstly, this paper concludes the background and significance of robust inverse mapping based on vanishing point, summarizes the research status of existing inverse perspective mapping algorithms and the existing problems.Secondly, this paper introduces the basic concepts and the mathematical principles of inverse perspective mapping. The inverse perspective mapping has widely been used in road traffic markings detection. The common inverse perspective mapping methods were divided into three categories based on the calculation of transformation parameters: the inverse perspective mapping based on the corresponding points, the inverse perspective mapping based on the transformation matrix and the inverse perspective mapping based on simplified camera model. It summarizes the three existing inverse perspective mapping methods and describes their advantages and disadvantages. Among them, this paper mainly describes the inverse perspective mapping based on the simplified camera model, because such an inverse perspective mapping method is simple and with high practical value.Thirdly, this paper presents a robust inverse perspective mapping method based on vanishing point detection. Vanishing point reflects the perspective structure of the image, the coordinates of the vanishing point can be used to calculate the camera pitch angle and yaw angle and provide real-time deflection angle parameters for inverse perspective mapping. It is more robustness to the downhill road and other unconventional road environment. At the same time, this paper improves image quality of an inverseperspective mapping result by the yaw angle correction. The proposed method firstly detected vanishing point, and then calculate the pitch angle and yaw angle according to the vanishing point, and finally get a top view of the road without perspective distortion through the inverse perspective mapping.Finally, the inverse perspective mapping method is used to detection and recognition of the arrow road markings. Arrow road makings is a kind of road traffic markings, which has an important indication function to regulate vehicle travel. Therefore, the detection and recognition of arrow road markings can provide important guide information to intelligent transportation systems. Firstly, it use the inverse perspective mapping method to eliminate the perspective influence of the road plane, and then detect the candidate region of arrow road markings, and then extract the shape feature by using shape contexts, at last the shape matching based on ant colony optimization algorithm is used to recognize the arrow road makings.Keywords:Vanishing Point Detection, Inverse Perspective Mapping, Intelligent Traffic System, Arrow Road Makings Detection and Recognition, ShapeMatching目录中文摘要 (I)英文摘要 (II)1 绪论 (1)1.1选题背景及研究意义 (1)1.2 研究现状及难点问题 (3)1.2.1 国内外研究现状 (3)1.2.2 研究的难点问题 (5)1.3 本文的主要工作及内容安排 (6)2 逆透视变换 (8)2.1 引言 (8)2.2逆透视变换的基本概念 (8)2.3 逆透视变换的数学模型 (9)2.3.1 智能车坐标系 (9)2.3.2 摄像机数学模型 (10)2.3.3 逆透视变换原理和性质 (13)2.4 逆透视变换应用 (14)2.4.1 道路交通标线检测和识别 (15)2.4.2 障碍物检测 (17)2.4.3 距离检测 (18)2.4.4 交通流检测 (19)2.5 逆透视变换实现 (20)2.5.1 基于矩阵变换的逆透视变换实现 (20)2.5.2 基于对应点变换的逆透视变换实现 (21)2.5.3 基于简化相机模型的逆透视变换实现 (22)2.6 本章小结 (24)3 基于消失点的鲁棒逆透视变换 (26)3.1 引言 (26)3.2 基于消失点的俯仰角和偏航角计算 (27)3.3 逆透视变换 (33)3.4 实验结果与分析 (36)3.5 本章小结 (39)4 基于IPM和形状匹配的导向箭头检测与识别 (40)4.1 引言 (40)4.2 基于逆透视变换的导向箭头检测 (42)4.2.1 基于逆透视变换的候选区域提取 (42)4.2.2 导向箭头检测 (44)4.3 基于形状匹配的导向箭头识别 (45)4.3.1 基于形状上下文的导向箭头形状特征提取 (46)4.3.2 基于蚁群优化的导向箭头形状匹配 (47)4.4 实验结果分析 (50)4.5 本章小结 (54)5 总结与展望 (55)致谢 (57)参考文献 (58)附录 (62)A. 作者在攻读学位期间内发表的论文目录 (62)B. 在攻读硕士学位期间参加的科研项目 (62)1 绪论本章首先介绍了基于消失点的鲁棒逆透视变换的选题背景及其研究意义,接着总结了现有逆透视变换算法的研究现状,并指出逆透视变换的难点问题,最后阐述了本文的主要研究工作和内容安排。