Spectral unmixing for hyperspectral image classification with an adaptive endmember selection

科研成果: 书/报告/会议事项章节会议稿件同行评审

2 引用 (Scopus)

摘要

Hyperspectral classification techniques are widely used for detailed analysis of the earth surface. However, mixed pixels caused by the relatively low spatial resolution of the imaging system are the big burden for traditional pure-pixel-hypothesis based hard classification methods. To address this problem, a novel method, which jointly uses soft classification and spectral unmixing, is proposed in this paper. The confusion matrix is exploited to determine the endmember set for each class. Then the generated endmember is adopted for spectral unmixing. The fractional abundance of training samples, which is generated from spectral unmixing, is utilized to optimize soft multinomial logistic regression classifier. The result of the optimized classifier will result in a more accurate confusion matrix. Thus, this procedure is executed iteratively to achieve required performance. Experimental results on synthetic and real hyperspectral data sets demonstrate the superiority of the proposed method for hyperspectral image classification.

源语言英语
主期刊名Intelligence Science and Big Data Engineering - 4th International Conference, IScIDE 2013, Revised Selected Papers
出版商Springer Verlag
359-367
页数9
ISBN(印刷版)9783642420566
DOI
出版状态已出版 - 2013
活动4th International Conference on Intelligence Science and Big Data Engineering, IScIDE 2013 - Beijing, 中国
期限: 31 7月 20132 8月 2013

出版系列

姓名Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
8261 LNCS
ISSN(印刷版)0302-9743
ISSN(电子版)1611-3349

会议

会议4th International Conference on Intelligence Science and Big Data Engineering, IScIDE 2013
国家/地区中国
Beijing
时期31/07/132/08/13

指纹

探究 'Spectral unmixing for hyperspectral image classification with an adaptive endmember selection' 的科研主题。它们共同构成独一无二的指纹。

引用此