基于非极大值抑制的圆目标亚像素中心定位  

Subpixel accuracy central location of circle target based on nonmaximum suppression

作　　者：王静[1,2] 王海亮[3] 向茂生[1] 韦立登[1] 刘忠胜[1] 

机构地区：[1]中国科学院电子学研究所微波成像技术国家级重点实验室,北京100190 [2]中国科学院研究生院,北京100190 [3]中国航天科工集团第二研究院二十三所,北京100854

出　　处：《仪器仪表学报》2012年第7期1460-1468,共9页Chinese Journal of Scientific Instrument

基　　金：国家863计划(2007AA120302);国家973计划(2009CB724000)资助项目

摘　　要：圆形目标在基于图像的测量系统中应用广泛,针对圆目标中心的高定位精度和快速提取的要求,提出了一种基于非极大值抑制的亚像素中心定位方法。该方法利用Sobel算子进行边缘检测,通过改进非极大值抑制方法获取连续细化的边缘,实现了像素级边缘定位,采用Zernike正交矩对边缘点进行亚像素级定位,并用最小二乘法进行二次拟合来获取精确的标志点的中心坐标。仿真图像和实测图像的实验结果验证了该方法的有效性和准确性,其定位精度可以达到0.02像素,通过测试算法的运行时间,证明该算法具有很好的实时性。Round target is frequently used in image measurement systems. To meet the requirements of high precision location and quick extraction of circle target center, A novel subpixel accuracy location method based on nonmaximum suppression was proposed. Sobel operator is used to detect the edge firstly. The nonmaximum suppression method is improved to obtain continuous and ideal circular edge, and the pixel accuracy location of circular edge is achieved. Zernike moment is used to carry out subpixel accuracy edge location of the edge. The circle target center is determined using the least mean-square algorithm with quadratic ellipse fitting. Experiment results on simulative and real images show the effectiveness and precision of the method ; the location accuracy can reach to 0.02 pixels, which meets the requirement of real-time measurement. Operation time test proves that the algorithm has good real time ability.

关 键 词：边缘检测 非极大值抑制 正交矩 椭圆拟合 中心定位 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]