作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (9): 1-3. doi: 10.3969/j.issn.1000-3428.2012.09.001

• 博士论文 •    下一篇

可变神经网络结构下的遥感影像光谱分解方法

李 熙1 ,石长民2,李 畅3,陈锋锐4,田礼乔1   

  1. (1. 武汉大学测绘遥感信息工程国家重点实验室,武汉 430079;2. 三亚市国土环境资源信息中心,海南 三亚 572000; 3. 华中师范大学城市与环境科学学院,武汉 430079;4. 河南大学环境与规划学院,河南 开封 475000)
  • 收稿日期:2011-08-22 出版日期:2012-05-05 发布日期:2012-05-05
  • 作者简介:李 熙(1982-),男,讲师、博士,主研方向:遥感影像处理;石长民,工程师、硕士;李 畅,讲师、博士;陈锋锐,博士研究生;田礼乔,讲师、博士
  • 基金资助:
    国家自然科学基金资助项目(41101413);高等学校博士 学科点专项科研基金资助项目(20110141120073);中央高校基本科 研业务费专项基金资助项目(904275839)

Spectral Unmixing Method of Remote Sensing Images in Variable Architecture of Neural Network

LI Xi 1, SHI Chang-min 2, LI Chang 3, CHEN Feng-rui 4, TIAN Li-qiao 1   

  1. (1. State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, China; 2. Sanya Land Environment and Resources Information Center, Sanya 572000, China; 3. College of Urban and Environmental Science, Huazhong Normal University, Wuhan 430079, China; 4. College of Environment and Planning, Henan University, Kaifeng 475000, China)
  • Received:2011-08-22 Online:2012-05-05 Published:2012-05-05

摘要: 多层感知神经网络(MLP)是主流的非线性分解方法,但是目前缺乏有效方法处理MLP分解结果中的丰度负值问题。为此,提出一种可变神经网络结构的方法,逐步去除负值丰度对应的端元,并调整相应的网络结构使之针对剩余的端元进行分解。通过武汉地区模拟TM遥感影像实验可以发现,该方法与传统MLP方法以及线性光谱分解方法的平均误差分别为0.077 7、0.081 9、0.094 3,说明该方法的分解精度高于其他2种分解方法,能克服丰度负值问题。

关键词: 遥感, 混合像元, 神经网络, 多层感知网络, 非负约束, 非线性光谱分解模型

Abstract: Spectral unmixing of remote sensing images is a hotspot in remote sensing field, and Multilayer Perception(MLP) neural network is a common nonlinear spectral unmixing algorithm. However, currently there is no effective way to deal with the negative abundances derived by the network. To solve this problem, a MLP neural network with variable architecture is proposed. By discarding endmembers with negative abundances, the MLP architecture is modified to unmix the rest endmembers, so a remote sensing image is finally unmixed. An experiment using a simulated TM image shows that the average errors of the proposed method, conventional MLP method and linear spectral unmixing model are 0.077 7, 0.081 9 and 0.094 3 respectively, thus the proposed method outperforms the other two. Therefore, the proposed method can overcome the negative abundance problem effectively.

Key words: remote sensing, mixed pixel, neural network, Multilayer Perception(MLP) network, nonnegative constraint, nonlinear spectral unmixing model

中图分类号: