基于SURF的遥感图像配准优化算法  

The Optimal Algorithm of Remote Sensing Image Registration Based on SURF

作　　者：张天翼 刘鹏[1] 史佳霖 毕誉轩 王彩霞[1] ZHANG Tianyi;LIU Peng;SHI Jialin;BI Yuxuan;WANG Caixia(School of Electronic and Information Engineering,Changchun University of Science and Technology,Changchun 130022)

机构地区：[1]长春理工大学电子信息工程学院,长春130022

出　　处：《长春理工大学学报（自然科学版）》2024年第1期98-104,共7页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：吉林省科技发展计划资助项目(20210201021GX)。

摘　　要：针对遥感图像配准中易出现的信息量巨大和误匹配问题,通过改进SURF算法,提出了一种用于遥感图像配准的优化算法。优化算法引入特征点图像熵,通过去除携带信息量不足的误选取特征点,提高了特征点选取以及后续特征点匹配的准确性。在特征点描述阶段,算法实现了描述子降维,并将降维后的特征点描述子与FLANN算法相结合,降低了算法特征点误匹配率,同时减少了因描述子降维而带来的影响。实验表明,该优化算法相较于传统SURF算法,待配准图像特征点选取率下降10.9%,特征点误匹配率下降17.5%,并得到精确的遥感图像配准效果。In response to the problem of large amounts of information and misalignment in remote sensing image registration,an optimized algorithm for remote sensing image registration is proposed by improving the SURF algorithm.The optimization algorithm introduces the concept of feature point image entropy to remove redundant feature points with limited information,thereby ensuring better accuracy of feature point selection and subsequent matching results.The algorithm achieves dimension reduction of the feature point descriptors at the description stage,and the reduced-dimensional feature point descriptors are combined with the FLANN algorithm to lower the feature point mismatch rate and mitigate the negative influence caused by dimension reduction.Experimental results demonstrate that the optimized algorithm significantly outperforms the traditional SURF algorithm in terms of feature point selection rate,with a reduction of 10.9%,and feature point mismatch rate,with a decrease of 17.5%,and delivers accurate registration results.

关 键 词：图像配准 SURF算法 图像熵 FLANN算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]