基于双平面靶标的多视觉传感器现场全局校准  

Global Calibration of Multi-sensor Vision System Based on Two Planar Targets

作　　者：刘震[1] 张广军[1] 魏振忠[1] 

机构地区：[1]北京航空航天大学精密光机电一体化技术教育部重点实验室,北京100083

出　　处：《机械工程学报》2009年第7期228-232,共5页Journal of Mechanical Engineering

基　　金：国家自然科学基金资助项目(50727502)

摘　　要：针对现有多视觉传感器全局校准方法多依赖外部辅助设备的问题,提出一种基于双平面靶标的多视觉传感器现场全局校准方法。以其中一个视觉传感器为基础视觉传感器,建立全局坐标系。根据基础视觉传感器与待校准视觉传感器之间的位置关系,调整固定两个平面靶标之间位置。将双平面靶标无约束移动多次(至少三次),每个视觉传感器可以"看"到对应的平面靶标。以两个平面靶标之间位置关系不变为约束条件,求解待校准视觉传感器坐标系到全局坐标系的转换矩阵,通过非线性优化方法求解转换矩阵在最大似然准则下的最优解。当视觉传感器多于两个时,通过两两视觉传感器校准的方式完成多视觉传感器现场全局校准。试验证明该方法灵活方便,切实可行。A global calibration method based on two planar targets is proposed to solve the problem that the global calibration of multi-sensor vision system depends on the external high-precision measurement equipment. The coordinate fi＇ame of one of the vision sensors, called the base vision sensor, is selected as the global coordinate fi＇ame. The relative position of two planar targets is fixed according to the relative position between the base vision sensor and the vision sensor to he calibrated. The two planar targets are placed in front of the two sensors several times （at least three times） with no restriction. Each sensor catches the image of the corresponding planar target each time. Using the invariance of the relative positions between the two planar targets, the transformation from the local coordinate frame of the vision sensor to be calibrated to the global coordinate frame can be solved. Maximum likelihood estimation of the transformation matrix is achieved through non-linear optimization algorithm. The global calibration can be realized by means of pair-wise calibration. Experimental results show that the method is easy, flexible and practical in the vision inspection applications.

关 键 词：机器视觉 多视觉传感器 双平面靶标 全局校准 柔性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]