基于直线夹角判决图像三点不共线的方法  

Noncollinear Judgment of 3-Point Based on Angles between Lines

作　　者：张永亮[1] 肖山竹[1] 

机构地区：[1]国防科学技术大学ATR实验室,湖南长沙410073

出　　处：《计算机仿真》2008年第6期169-172,共4页Computer Simulation

基　　金：国家高技术研究发展计划(863计划)(2006AA801412)

摘　　要：机器人自主导航是计算机视觉几何估计问题研究的重要应用领域之一。特征点对应的射影变换独立性则是计算机视觉几何估计算法中的基本要求,RANSAC算法常用的计算机视觉几何估计算法,具有鲁棒性好的优点,在进行目标结构和摄像机运动估计前对任意三点不共线进行判决可以提高RANSAC算法的收敛速度。分析了图像特征点提取噪声模型,提出了基于直线夹角判决三点不共线的方法,设计了仿真实验,根据多次实验结果观察的统计规律性获取了判决的夹角阈值,对工程应用具有一定的指导价值,据作者所知没有类似的文献发表。Automatic navigation of robot is one of the most important application fields of geometry estimation in computer vision. Projective independent of feature - points＇ correspondences is one of the basic requirements of the algorithms of geometry estimation in computer vision. And RANSAC （ RANdom SAmpling Consensus）, which is robust, is one of the most widely used algorithms of geometry estimation. The processing of noncollinear judgment of 3 -point before SAM （Structure and Motion） can increase the convergence speed of RANSAC algorithms. The paper analyses the noise model of abstracting feature - points, proposes a method of noncollinear judgment of 3 - point based on angles between lines, designs the experiment and gains the angular threshold through a lot of experiments and their statistic property of results. It is of great practical importance and there is no similar literature about it until now.

关 键 词：直线夹角 不共线判决 几何估计 机器人导航 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]