浅谈SIFT算法对HJ-1B影像自动配准研究  

Discussion on Automatic Registration of HJ-1B image by SIFT algorithm

作　　者：刘志强 孔玉霞 牟洪洲 路岩 王博 LIU Zhi-qiang;KONG Yu-xia;MU Hong-zhou;LU Yan;WANG Bo(Liaoning Institute of Science and Technology, Benxi, Liaoning, 117004, China)

机构地区：[1]辽宁科技学院资源与土木工程学院,辽宁本溪117004

出　　处：《辽宁科技学院学报》2019年第1期34-36,54,共4页Journal of Liaoning Institute of Science and Technology

基　　金：2018年辽宁科技学院大学生创新创业训练计划项目"基于AROP的HJ-1B卫星图像的自动配准研究"(201811430152)

摘　　要：实现影像间自动配准的关键在于快速提取足够数量且精度高的匹配点。提取特征点的方法主要有基于特征和基于灰度两种。经过大量的实验证明,sift算法提取特征点,从粗略的提取特征点,再通过四步对已提取的特征点进行提纯,最后利用精度较高的匹配点进行计算影像间的变换参数,实现配准的自动化。The key for realizing automatic image registration is to quickly extract a sufficient number of matching points with high precision.There are two methods to extract the feature points: based on feature and based on gray scale. Through a large number of experiment, it is proved that the shift algorithm is the best one. After roughly extracting feature points, purify the extracted feature points through four steps, Finally calculate the transformation parameters between images by using the matching points with high precision, so as to achieve automatic registration.

关 键 词：SIFT算法 自动配准 配准精度 

分 类 号：G64[文化科学—高等教育学] TF-4[文化科学—教育学]