作　　者： ; ; ;

机构地区： 中国科学院沈阳自动化研究所机器人学国家重点实验室

出　　处： 《计算机学报》 2008年第3期493-501,共9页

摘　　要： 基于旋转体的摄像机定位是单目合作目标定位领域中的涉及较少并且较为困难的一个问题,传统的基于点基元、直线基元及曲线基元的定位方法在用于旋转体定位过程中都存在相应的问题.文中设计了一种由4个相切椭圆构成的几何模型,该模型环绕于圆柱体表面,利用二次曲线的投影仍然是二次曲线的特性和椭圆的相应性质能够得到唯一确定模型位置的3个坐标点,从而将旋转体定位问题转化为P3P问题.在对P3P的解模式区域进行分析后,推导了根据模型上可视曲线的弯曲情况来确定P3P问题解模式的判别方法,并给出证明过程.仿真实验表明了这种模型定位方法的有效性.最后利用这个模型引导机械手完成目标定位的实验. Camera pose determination for body of revolution with single camera is a problem that is referred little and more puzzled. Traditional pose estimation methods using point features, straight line features and conics features have corresponding problems when they are applied to pose computation based on body of revolution. This paper presents a geometric model composed of four tangential ellipses that surround the surface of a cylinder. By using the preserved characteristic of conics and the corresponding characteristic of ellipse, three control points which can determine the pose of the model uniquely are obtained, thus the current problem of is converted into the problem of P3P. At the same time, a judgment method is developed to get the solution mode of P3P according to the curving style of eyeable arc on the model based on the analysis of the solution mode region of P3P. After that the proving process is presented. The effectiveness of the location method based on model is confirmed by simulation. At last, the model is used to lead a manipulator to locate the object in an experiment.

关 键 词： 旋转体 问题 多解现象 唯一解

领　　域： [自动化与计算机技术] [自动化与计算机技术]