单目视觉位姿测量方法及数字仿真  

Position and Orientation Measurement Method with Monocular Vision and Digital Simulation

作　　者：张子淼[1] 王鹏[1] 孙长库[1] 

机构地区：[1]天津大学精密测试技术及仪器国家重点实验室,天津300072

出　　处：《天津大学学报》2011年第5期440-444,共5页Journal of Tianjin University(Science and Technology)

基　　金：国家科技重大专项支持资助项目(2009ZX04014-092)

摘　　要：对基于点特征的单目视觉定位算法进行研究,提出了一种基于5点透视投影数学模型的单目视觉位姿测量算法,该算法使用不共面定位特征点,根据5个特征点在摄像机坐标系与世界坐标系下相对位置关系相同来计算5个特征点在摄像机坐标系下的坐标,进而求出物体的转动角度.同时还构建了一套基于OpenGL图形引擎的数字仿真系统,完成了物体位姿的仿真计算,实现对算法定位精度的评价,仿真实验结果表明,在理想状况下,该算法的定位精度达到了±0.01°,为单目视觉物体位姿测量提供了一种具有较高精度的方法.The monocular vision positioning algorithm which is based on the point characteristics is studied.The algo-rithm which is based on the mathematical model of 5-point perspective projection is proposed.The algorithm uses non-coplanar feature points and calculates the coordinates of five feature points in the camera coordinate system ac-cording to the same relative position of the five feature points of the object surface in the camera coordinate system and in the world coordinate system.So the object rotation angle can be calculated.A set of digital simulation system based on OpenGL graphics engine is constructed to simulate the algorithm for completing the specific process of posi-tioning algorithm simulation and verifying the accuracy of the algorithm.Simulation results show that the positioning accuracy of the algorithm reached ±0.01°in the ideal case.The algorithm offers a new method with high precision for the measurement of the position and orientation of the object.

关 键 词：单目视觉 物体位姿 五点透视 数字仿真 

分 类 号：O29[理学—应用数学] TB11[理学—数学]