首页 | 本学科首页   官方微博 | 高级检索  
     

高光谱图像分类的研究进展
引用本文:闫敬文,陈宏达,刘蕾. 高光谱图像分类的研究进展[J]. 光学精密工程, 2019, 27(3): 680-693. DOI: 10.3788/OPE.20192703.0680
作者姓名:闫敬文  陈宏达  刘蕾
作者单位:汕头大学工学院电子系,广东汕头,515063;汕头大学医学院,广东汕头,515063
基金项目:国家自然科学基金资助项目(No. 61672335,No. 61601276);广东省自然科学基金资助项目(No. 2016A030310077)
摘    要:高光谱图像分类是利用高光谱数据图谱合一且光谱信息丰富的特点,对图像中的每个像素进行分门别类,以达到对地物目标进行高精度分类和自动化识别的目的,是对地观测的重要组成部分。在分析高光谱图像特点的基础上,本文从普通机器学习和深度学习这两方面对高光谱图像像素级分类的研究进展及效果进行总结、评述和比较,通过具体实验的结果对比,直观地展现各种算法的优劣。针对高光谱分类问题,本文从两个方面对今后的研究方向及发展前景进行了分析和展望。一方面,在算法研究上,高光谱图像分类算法可在保证分类精度的前提下降低算法的复杂度,利用多源遥感数据、多特征综合、多尺度复合,提升小样本、少参数分类模型的分类精度,适应智能化、快速化高光谱遥感对地观测的发展要求;另一方面要紧密结合市场应用需求,重视高光谱图像在实际中的应用,研究具有市场竞争力的高效分类算法,提升高光谱图像分类在遥感技术应用领域的竞争力。

关 键 词:高光谱图像  像素级分类  机器学习  深度学习
收稿时间:2018-10-30

Overview of Hyperspectral Image Classification
YAN Jing-wen CHEN Hong-da LIU Lei. Overview of Hyperspectral Image Classification[J]. Optics and Precision Engineering, 2019, 27(3): 680-693. DOI: 10.3788/OPE.20192703.0680
Authors:YAN Jing-wen CHEN Hong-da LIU Lei
Affiliation:1. Department of Electronics, Shantou University, Shantou 515063, China;2. Medical College, Shantou University, Shantou 515063, China
Abstract:Hyperspectral image classification is to classify every pixel of the image by the combination of hyperspectral data atlas and rich spectral information, which can be used to achieve high-precision classification and automatic recognition of ground objects. It is an important part of earth observation. Based on the analysis of the characteristics of hyperspectral images in two aspects of general machine learning and depth learning,the research progress and compares effects of pixel-level classification of hyperspectral images are summarized and commented. By comparing the results, the advantages and disadvantages of various algorithms are visually displayed. Research directions and development prospects of hyperspectral images classification are analyzed and prospected from two aspects. On the one hand, it is about algorithm research. Hyperspectral classification algorithm can guarantee the accuracy of classification to reduce the algorithm complexity by using multi-source remote sensing data, multi-feature and multi-scale composite, improve the classification accuracy of small sample and less parameters of the classification model, adapt the intelligent and fast development requirements of earth observation. On the other hand, market application needs to be closely integrated. Attention should be paid to the application of hyperspectral images in practice and efficient classification algorithm with market competitiveness should be researched to enhance the competitiveness of hyperspectral image classification in remote sensing applications.
Keywords:hyperspectral image   pixel-level classification   machine learning   deep learning
本文献已被 CNKI 万方数据 等数据库收录!
点击此处可从《光学精密工程》浏览原始摘要信息
点击此处可从《光学精密工程》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号