一种抗噪的高精度圆形阵列靶标特征提取算法  

An Anti-noise Precise Algorithm for Feature Extraction of Circle-grid Target

作　　者：凌乐[1] 陈远强 周东 LING Le;CHEN Yuanqiang;ZHOU Dong(Central Research Academy of Dongfang Electric Corporation, 611731, Chengdu, China)

机构地区：[1]东方电气集团中央研究院,成都611731

出　　处：《东方电气评论》2017年第3期1-7,共7页Dongfang Electric Review

摘　　要：实验发现目前普遍采用的棋盘格靶标角点提取算法对噪声十分敏感,其特征提取耗时及检测误差随着噪声均值增加而明显增大,无法满足高噪环境下的使用需求。本文使用圆点阵列取代棋盘格作为图像靶标,并设计了一种具有良好抗噪性的圆点阵列靶标的特征提取算法。该算法首先采用多级阈值对靶标图像进行二值化,然后对二值化图像序列进行连通域检测,通过矩方法对每一个连通域的形态进行分析计算,排除误检点,采用最小外接圆法优化已检圆的特征位置,最后根据靶标的尺寸信息,从所有已检圆点中,搜寻出靶标的所有特征圆。实验结果表明该算法在同等高斯噪声下特征提取误差较棋盘格小一个数量级,且常用的滤波算法在不会明显影响其特征提取精度的前提下能有效提高检测速度。Testings reveal that the current algorithm for corner extraction of checkerboard target is sensitive to noise,therefore,a circle grid board is used for image target,and an efficient algorithm designed for feature extraction with good noise immunity.First,a multi-thresliold issetfor imagebinarization.Then,theconnecteddomainisdetectedforeachbinaryimage,afterthat,themoment methodis used to calculate the form of each connected domain to exclude the wrong features,and to optimize the location of circle through the minimum circumscribed circle method.Finally,according the board size,the expected cirlce grid is found out from all remaining connecteddomain.The testings indicate that this algorithm has better noise immunity than checkerboard corners extraction bothin detection efficiency and accuracy.And the Gauss image filter will not reduce the accuracy dramaticlly,but raise the speed of detection efectivel.

关 键 词：特征提取 矩方法 抗噪性 图像滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]