基于单幅图像大尺寸实际平面上的光点定位方法  

3D reconstruction of feature point on large scale object surface from a single image

作　　者：杨明[1] 仲小清[1] 霍炬[1] 

机构地区：[1]哈尔滨工业大学控制与仿真中心,黑龙江哈尔滨150080

出　　处：《应用光学》2011年第3期430-435,共6页Journal of Applied Optics

基　　金：黑龙江省杰出青年科学基金(JC200606);哈尔滨工业大学优秀青年教师培养计划(HITQNJS2007021)

摘　　要：为满足某地面实验中视觉测量系统的研制需求,研究了一种基于单幅图像的大尺寸实际平面上光点的定位方法。通过建立关于实际平面上光点3D坐标的数学模型,结合利用图像确定的通过光点的空间光束,实现了由单幅图像确定光点的3D坐标。依据具体特点,把实际平面分为高阶曲面型、分块平面型和分块曲面型3种类型,并分别给出了相应类型的光点定位方法。仿真实验验证了所提出方法的正确性,并比较分析了方法的精确性;实际试验数据表明,所提出方法可以满足尺寸为8 000 mm×8 000 mm的大型实际平面上光点的定位需求,定位误差小于1 mm。For the design and development of a vision sensor for the ground test,a method of 3D reconstruction of feature point on large scale object surface from a single image is proposed.By establishing the math model of the 3D coordinate of feature point on the object surface and using the spatial ray through feature point,the 3D coordinate of the feature point can be determined using sigle image.According to the characteristics,the object surface can be classified into three types,high order surface type,block plane type and block surface type,while the corresponding location methods are introduced.Simulation experiment is conducted to compare the accuracy of three different 3D reconstruction methods,and the proposed method performs the best.By the measurement precision of 1 mm in the range of 8000mm×8000mm,it is proved that the proposed method is suitable for 3D reconstruction of feature point on large scale object surface.

关 键 词：机器视觉 大尺寸光点定位 实际平面点 单幅图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]