基于迭代逼近的圆标志精密定位算法研究  

Research on the Precise Location Algorithm of Circle Target Based on Iteration and Approximation

作　　者：周年芳[1,2] 孙久运 王小兵[3] 柏林森[2] 

机构地区：[1]中国矿业大学江苏省资源环境信息工程重点实验室,徐州221116 [2]中国矿业大学环境与测绘学院,徐州221116 [3]中钢集团马鞍山矿山研究院有限公司,马鞍山243000

出　　处：《科学技术与工程》2014年第15期72-78,共7页Science Technology and Engineering

基　　金：国家自然科学基金(41171343);江苏高效优势学科建设工程项目(SZBF2011-6-B35)资助

摘　　要：目前圆标志定位主要通过拟合边缘像素获取中心位置,理想情况可达亚像素级精度;但计算复杂且受噪声影响大。对此,引入一种无需圆拟合,在边缘提取与粗定位基础上利用圆标志边缘像素迭代逼近圆心的高精度定位方法;并着重探究迭代中心点误差随叠加噪声与迭代次数不同而存在的影响规律。研究结果表明:在叠加相同水平噪声而迭代不同次数情况下,迭代中心点误差呈指数函数分布;在迭代相同次数而叠加不同水平噪声情况下,其呈傅里叶函数分布;且圆标志x与y方向误差分布模型相似,进一步说明通过该类模型可对圆标志在不同噪声下的定位精度进行较为精确地估计。最后实际标志成像计算结果表明定位精度达到0.02像素。Currently location algorithm of circle target is mainly ellipse-fitting by edge pixels,under ideal condition,it can obtain sub-pixel precision. However,it has complex computations and huge influences from noise. So a precise center location algorithm,iteration and approximation through edge pixels based on edge extraction and coarse location without ellipse-fitting,is introduced. And analysis of laws between center error and noise and iterative times are emphasized. It turns out that center error present as exponential function in the case of same noise and different iterative times while it presents as fourier function in the case of same iterative times and different noise. The similarity between error model of x and y further proves that can precisely forecast location error of circle target under various noise conditions with the function model above. At last,tests demonstrate that the center is localized in 0. 02 pixel.

关 键 词：圆标志定位 亚像素 迭代逼近 函数模型 多结构多尺度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]