用于视觉传感器标定的鲁棒性靶标定位算法  

Robust pattern locating algorithm for calibration of vision sensor

作　　者：崔海华[1] 廖文和[1] 程筱胜[1] 戴宁[1] 傅士强[1] 耿海宁 

机构地区：[1]南京航空航天大学机电学院,江苏南京210016 [2]解放军总装备部江阴远望测量基地,江苏江阴214431

出　　处：《传感器与微系统》2009年第9期62-65,69,共5页Transducer and Microsystem Technologies

基　　金：国家"863"计划资助项目(2005AA420240);教育部博士点基金资助项目(20070287055);江苏省数字化制造技术重点建设实验室开放课题(HGDML-0609)

摘　　要：为了满足高精度机器视觉传感器的标定,提出了一种新的靶标识别定位算法。算法的实现包括靶标模式的检测和匹配两部分,其中,基于矩的靶标模式检测方法使得检测精度与模式形状无关,克服了摄影变换后圆形靶标模式的非圆性缺点,保证了亚像素级的检测精度;提出的种子扩散算法能根据靶标邻域信息自动调整扩散方向和度量信息,对于任意位姿的靶标图像都能保证与三维靶标实现稳定、准确、自动的匹配。对新设计的靶标进行了识别定位实验,结果表明:算法具有高精度、高正确定位重复率、全自动和快速性等优点。A new pattern locating algorithm is presented for high precision calibration of the machine vision sensor. This locating algorithm includes accurate target detection and pattern matching. The developed subpixel estimation method based on moment is immune to the pattern shape, so it overcomes the non-circular disadvantage of circular target patterns by photography transformation, and ensures the detection accuracy. The proposed matching algorithm can adaptively adjust the matching direction and distance metrics according to the information of neighbor targets. This matching algorithm is stable, automated for any attitude of the pattern image. The new locating algorithm is used to recognize the new design pattern, the experiments show the locating algorithm has high precision, high repetition rate of correct locating, automatic and fast features.

关 键 词：机器视觉 靶标识别 靶标匹配 系统标定 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]