摘要
针对传统Zernike矩亚像素边缘检测算法中奇数尺寸模板卷积的不对称性,在分析卷积不对称性和卷积窗口中心局限性对检测精度影响的基础上,提出一种改进的Zernike矩亚像素边缘检测算法。算法采用像素级边缘检测算子进行初定位,获取像素级边缘点;利用小尺寸奇数模板求解Zernike矩进行二级初定位,获取亚像素级边缘点;分析二级初定位边缘点所在区间;根据所在区间,选取偶数模板或者奇数模板求解Zernike矩进行精定位,获取更精确的亚像素级边缘点。为了验证算法的有效性,分别对模拟图像和实际图像进行亚像素边缘检测实验。根据齿轮的边缘特性,通过最小二乘拟合法评价算法对齿轮图像的检测效果。实验结果表明,该算法具有绝对定位精度高、抗噪性能强、曝光敏感度好的优点。
This paper analyzes the influence of the asymmetry on the odd template convolution. The limitation in the center of convolution window on the detection accuracy of traditional Zernike moment sub-pixel edge detection algorithm is discussed. An improved Zernike moment sub-pixel edge detection algorithm is proposed. The algorithm utilizes the pixel level operator to obtain the pixel level edge points in the initial location. Then,sub-pixel level edge points are obtained by using small odd template to solve Zernike moment in the secondary initial positioning. The interval in the secondary initial positioning edge points is determined. Based on this interval,the even templates or odd templates are utilized to solve the Zernike moments for fine positioning to obtain more accurate sub-pixel edge positions.Experiments comparison between the proposed algorithm and other algorithms have been realized to verify its performance based on the simulative and real images. The detection effectiveness for gear image is evaluated by the least-square fitting. Experimental results show that the proposed algorithm has high precision of absolute positioning,strong anti-noise performance,and better exposure sensitivity.
作者
刘明佩
朱维斌
叶树亮
Liu Mingpei;Zhu Weibin;Ye Shuliang(Institute of Industry and Trade Measurement Technique,China University of Metrology,Hangzhott 310018,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2018年第8期259-267,共9页
Chinese Journal of Scientific Instrument
基金
国家质量监督检验检疫总局科技计划(2014QK238)
国家重大科学仪器设备开发专项(2013YQ470767)资助
