摘要
数字图像相关法的初始值对算法的计算效率和求解精度都有较大的影响,为此提出一种利用密集特征匹配求取初始值的算法。使用AKZAE算子对特征点进行检测,使用Daisy描述符对特征点进行描述,再采用网格运动统计(GMS)算法对特征点进行筛选以求取初始值,最终将初始值代入反向组合高斯牛顿(IC-GN)法中迭代求解亚像素位移。与SIFT(Scale Invariant Feature Transform)和SURF(Speeded-Up Robust Features)算法相比,AKAZE算子提高定位的准确性,而且计算效率更高,是一种兼顾速度与稳定性的特征点检测算法;Daisy描述符是一种高效的稠密特征提取描述符,相比于其他描述符能够实现更加密集的特征提取。
The initial value of the digital image correlation method has a great influence on the calculation efficiency and solution accuracy of the algorithm.For this reason,an algorithm using dense feature matching to obtain the initial value is proposed.The AKZAE operator is used to detect the feature points,the Daisy descriptor is used to describe the feature points,and then the grid motion statistics(GMS)algorithm is used to filter the feature points to obtain the initial value,and finally the initial value into the reverse combined Gaussian in the Newton(IC-GN)method is substituted,the sub-pixel displacement is solved iteratively.Compared with SIFT(Scale Invariant Feature Transform)and SURF(Speeded-Up Robust Features)algorithms,the AKAZE operator improves the accuracy of positioning and has higher computational efficiency.It is a feature point detection algorithm that takes into account both speed and stability.The Daisy descriptor is an efficient dense feature extraction descriptor,which can achieve denser feature extraction compared to other descriptors.
作者
谭芳喜
肖世德
李晟尧
周亮君
Tan Fangxi;Xiao Shide;Li Shengyao;Zhou Liangjun(School of Mechanical Engineering,Southuest Jiaotong University,Chengdu,Sichuan 610031,China;Techmology and Equipment of Rail Transit Operation and Maintenance Key Laboratory of Sichuan Province,Chengdu,Sichuan 610031,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2021年第16期382-390,共9页
Laser & Optoelectronics Progress
基金
四川省应用基础研究基金(2014JY0212)。
关键词
测量
数字图像相关
密集特征匹配
网格运动统计
反向组合高斯牛顿法
measurement
digital image correlation
dense feature matching
grid-based motion statistics
inverse compositional Gauss-Newton method
