首页 >  , Vol. , Issue () : -

摘要

全文摘要次数: 1097 全文下载次数: 978
引用本文:

DOI:

10.11834/jrs.20211358

收稿日期:

2021-05-27

修改日期:

2021-09-30

PDF Free   EndNote   BibTeX
遥感影像深度学习配准方法综述
李星华1, 艾文浩1, 冯蕊涛2, 罗少杰3
1.武汉大学 遥感信息工程学院 武汉;2.陕西师范大学 地理科学与旅游学院 西安;3.国网浙江省电力有限公司杭州供电公司 杭州
摘要:

遥感影像配准是指通过几何变换使两景或多景影像空间位置对齐的过程,是影像融合、变化检测、农业监测等应用的重要预处理步骤。近年来,深度学习引起了人们的广泛关注,并在遥感影像配准中成功应用。本文在简要介绍传统遥感影像配准方法的基础上,重点分析了深度学习在基于特征的配准方法、基于区域的配准方法两方面取得的重要进展,分享了用于遥感影像配准的公开数据集,并总结了深度学习在遥感影像配准中的机遇与挑战。

Survey of remote sensing image registration based on deep learning
Abstract:

Remote sensing image registration is the process of spatial alignment of two or more images through geometric transformation. It is an important preprocessing operation for image fusion, change detection, agricultural monitoring and other remote sensing applications. Considering that remote sensing image has the characteristics of large-scale change, complex ground covers and imaging modalities, although a large number of registration methods have been developed, there is still a lack of methods which can be widely used in different scenarios. Therefore, the research on registration algorithm with high efficiency, high robustness, high precision and wide applicability is of great significance. In recent years, deep learning which achieves great success in the field of natural image and medical image registration, provides a new way for remote sensing image registration. Firstly, we introduced two kinds of traditional registration methods, and analyzed the advantages and disadvantages of area-based and feature-based registration methods in details from the aspects of registration accuracy, efficiency and algorithm robustness. Generally speaking, there are two main problems in traditional methods: poor applicability and insufficient utilization of the deep semantic information of the image. Secondly, we focused on the important progress of deep learning in area-based registration method and feature-based registration method. According to the specific application purpose of deep learning, we made a more detailed division of the above two methods, and summarized the advantages and disadvantages of the existing research. In addition, considering the importance of datasets for deep learning, we sorted and shared some public datasets for remote sensing image registration. Thanks to the great progress of earth observation technology, more and more remote sensing images are put into application. Image registration is the key step of remote sensing image preprocessing and the basic research content of quantitative remote sensing analysis. In recent years, the research on remote sensing image registration algorithm based on deep learning has shown an increasing trend, but it is still in the early stage, and the framework is not mature. It mainly includes but is not limited to the following shortcomings: (1) lack of open source standard datasets; (2) difficult to apply to large-scale remote sensing images; (3) insufficient utilization of geospatial information and spectral information of remote sensing images; (4) long training time and the large computing overhead. From the perspective of data and methods, we looked forward to the application of deep learning in the field of remote sensing image registration, and put forward four main research directions: (1) remote sensing image registration datasets; (2) registration methods based on hybrid model; (3) registration methods based on different neural networks; (4) training strategy based on small samples.

本文暂时没有被引用!

欢迎关注学报微信

遥感学报交流群 分享按钮