Please wait a minute...
img img
高级检索
地球科学进展  2003, Vol. 18 Issue (3): 345-350    DOI: 10.11867/j.issn.1001-8166.2003.03.0345
研究论文     
ASTER数据的自组织神经网络分类研究
哈斯巴干,马建文,李启青
中国科学院遥感应用研究所,北京 100101
STUDY ON ASTER DATA CLASSIFICATION USING SELF-ORGANIZING NEURAL NETWORK METHOD
Hasi Bagan,Ma Jianwen, Li Qiqing
The Institute of Remote Sensing Applications, CAS, Beijing 100101,China
 全文: PDF(189 KB)  
摘要:

传统的遥感数据分类方法大多基于统计学的参数估计,假设数据分布服从高斯正态分布。神经网络方法无需参数估计和统计假设,因而,近来越来越多地应用于遥感数据分类之中。介绍了基于聚类分析的自组织特征映射分类方法。ASTER卫星数据是新型遥感数据,包括 3个15 m分辨率波段和 3个30 m分辨率的短波红外波段。选择北京地区的ASTER数据作为方法实验数据,首先对数据进行了小波融合,然后进行了土地覆盖类型的自组织特征映射神经网络分类研究,把研究结果同最大似然判别法得到的分类结果进行了比较,分类精度比最大似然判别法总体提高了9%。

关键词: 分类小波融合自组织特征映射神经网络    
Abstract:

    The assumption of statistical model is not needed for Neural Networks (NN) while most traditional classification method for remote sensing data assumed normal distribution model. More and more NN application cases have been found in remote sensing data classification. In this paper, we proposed a method of Kohonen Self-organizing feature map based on clustering analysis. ASTER data is a new remote sensing data, which includes 3 bands of 15 m resolution and 3 bands of 30m resolution. ASTER data of Beijing have been chosen for our research. The land cover classification result in neural networks method has been shown in this paper after wavelet fusion of data. The classification has 9% of accuracy ratio more than MLH classification.
    The idea of neural networks came from the basic structure of functioning of the human brain. In the modern field of science and engineering, the neural networks have strengthened their importance with numerous applications ranging from pattern recognition, fields of classification etc. There are different kinds of the neural networks available depending on the task to be performed. In this study the Kohonen self-organized network is used. There are 6 notes in import layer of the structure of Kohonen self-organized network and ASTER data bands 1,2,3N,5,7,9 corresponding to one note in import layer. Output layer has the structure of 25×25 neural notes. Learning speed α starting value is 0.9, α reduced to 0.001 stopped with net calculation processing. Maximum circulation time is 2 500.  
    ASTER is the only instrument to fly on the EOS AM-1 plate form that will acquire high-resolution image. The primary goal of the ASTER mission is to obtain high-resolution image data in 15 channels over targeted areas of the Earth's surface, as well as black-and-white stereo images, with a revisit time between 4 and 16 days. Band 1、2 are visible bands, band 3N,3B are near inferred bands, the resolution is 15 m; Band from 4 to 9 are group of  short wave inferred bands, theresolution is 30 m; Band from 10~14 are thermal bands, the resolution is 90m. With ASTER's merits earth scientists to address a wide range of globule-change topics. In the paper we introduce Kohonen self-organized network in classification of land cover in Beijing area in 2001 by using ASTER data.

Key words: Classification    Wavelet fusion    Self-organizing feature map    Neural networks.
收稿日期: 2002-10-18 出版日期: 2003-06-01
:  TP183  
基金资助:

国家863计划项目“基于SIG框架的数字城市服务系统与示范”(编号:2002AA134030);863-103项目“遥感图像处理平台”(编号:2002AA133030)资助.

通讯作者: 哈斯巴干     E-mail: hasibagan@263.net
作者简介: 哈斯巴干(1967-),男,内蒙古翁牛特旗人,博士研究生,主要从事遥感数据模型研究.E-mail: hasibagan@263.net
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章  
马建文
李启青
哈斯巴干

引用本文:

哈斯巴干,马建文,李启青. ASTER数据的自组织神经网络分类研究[J]. 地球科学进展, 2003, 18(3): 345-350.

Hasi Bagan,Ma Jianwen, Li Qiqing. STUDY ON ASTER DATA CLASSIFICATION USING SELF-ORGANIZING NEURAL NETWORK METHOD. Advances in Earth Science, 2003, 18(3): 345-350.

链接本文:

http://www.adearth.ac.cn/CN/10.11867/j.issn.1001-8166.2003.03.0345        http://www.adearth.ac.cn/CN/Y2003/V18/I3/345

[1] Zhang Xiaocan,Huang Zhicai,Zhao Yuanhong.Remote Sensing Digital Image Processing[M]. Hangzhou:ZheJiang University Press,1997.222-232.[张孝灿,黄智才,赵元洪. 遥感数字图像处理[M]. 杭州: 浙江大学出版社, 1997.222-232.]

[2] Luo Jiancheng,Wang Qinmin,Ma Jianghong,et al. The EM-based maximum likelihood classifier for remotely sensed data[J]. Acta Geodaetica et Cartographica Sinica,2002,31(2):234-239.[骆剑承,王钦敏,马江洪,.遥感图像最大似然分类方法的EM改进算法[J].测绘学报,2002,31(2):234-239.]

[3] Sun Jiabing, Shu Ning, Guan Zequn. Remote Sensing Theory,Method and Application[M].Beijing: Surveying and Mapping Press, 1997.176-242.[孙家炳,舒宁,关泽群.遥感原理、方法和应用[M]. 北京:测绘出版社,1997.176-242.]

[4] Kohonen T. Self-organizated formation of topologically correct feature maps[J]. Biological Cybernetics, 1982, 43: 59-69.

[5] C Y Ji. Land-use classification of remotely sensed data using kohonen self-organizing feature map neural networks[J]. Photogrammetric Engineering & Remote Sensing, 2000, 66(12): 1 451-1 460.

[6] Cui Jintai. An Introduction to Wavelets[M]. Xi'an: Xi'an Jiaotong University Press, 1995.198-242.[崔锦泰.小波分析导论[M].西安:西安交通大学出版社,1995.198-242.]

[7] Mallat S G. A theory of multiresolution signal decomposition the wavelet representation[J]. IEEE Transaction on Pattern Analysis and Machine Intelligence, 1989,11(7): 674-693.

[8] Jorge Nunez, Xavier Otazu, Octavi Fors, et al. Multiresolution-based image fusion with additive wavelet decomposition[J]. IEEE Transactions on Geoscience and Remote Sensing, 1999,37(3):1 204-1 211. 

[9] Hasibagan,Ma Jianwen,Li Qiqing,et al.Wavelet high frequency substitution fusion method[J]. Journal of Image and Graphics,2002,7A(10):1 012-1 016.[哈斯巴干,马建文,李启青,.小波局部高频替代融合方法[J].中国图象图形学报,2002,7A(10):1 012-1 016.]

[10] Kohonen T. The self-organizing map[J]. Proceedings of the IEEE, 1990, 78(9): 1 464-1 480.

[11] Yuan Zengren. Artificial Neural Network and Application[M]. Beijing:Tsinghua University Press,1999.[袁曾任. 人工神经元网络及其应用[M]. 北京: 清华大学出版社,1999.]

[12] Li Deren. Discussion on Earth observation and geographic information system[J]. Advances in Earth Sciences, 2001, 16(5):689-703.[李德仁.对地观测与地理信息系统[J].地球科学进展,2001,16(5):689-703.]

[1] 姜波, 李明, 屈争辉, 刘杰刚, 李伍. 构造煤研究现状及展望[J]. 地球科学进展, 2016, 31(4): 335-346.
[2] 陈科贵, 吴刘磊, 陈愿愿, 王刚. 基于支持向量机的川中杂卤石分类识别研究[J]. 地球科学进展, 2016, 31(10): 1041-1046.
[3] 于文涛, 李静, 柳钦火, 曾也鲁, 尹高飞, 赵静, 徐保东. 中国地表覆盖异质性参数提取与分析[J]. 地球科学进展, 2016, 31(10): 1067-1077.
[4] 陈科贵, 陈旭, 张家浩. 复合渗透率测井评价方法在砂砾岩稠油油藏的应用*——以克拉玛依油田某区八道湾组为例[J]. 地球科学进展, 2015, 30(7): 773-779.
[5] 李育, 朱耿睿. 三大自然区过渡地带近50年来气候类型变化及其对气候变化的响应[J]. 地球科学进展, 2015, 30(7): 791-801.
[6] 邹学勇, 张春来, 程宏, 亢力强, 吴晓旭, 常春平, 王周龙, 张峰, 李继峰, 刘辰琛, 刘博, 田金鹭. 土壤风蚀模型中的影响因子分类与表达[J]. 地球科学进展, 2014, 29(8): 875-889.
[7] 李艳雯, 邢喆, 李四海, 樊妙. 基于海底地名命名的海底地理实体分类进展[J]. 地球科学进展, 2014, 29(6): 756-764.
[8] 孟春雷. 城市地表特征数值模拟研究进展[J]. 地球科学进展, 2014, 29(4): 464-474.
[9] 王卷乐, 林海, 冉盈盈, 周玉洁, 宋佳, 杜佳. 面向数据共享的地球系统科学数据分类探讨[J]. 地球科学进展, 2014, 29(2): 265-274.
[10] 张盼盼, 刘小平, 王雅杰, 孙雪娇. 页岩纳米孔隙研究新进展[J]. 地球科学进展, 2014, 29(11): 1242-1249.
[11] 王志慧, 刘良云. 黑河中游绿洲灌溉区土地覆盖与种植结构空间格局遥感监测[J]. 地球科学进展, 2013, 28(8): 948-956.
[12] 李启权,王昌全,岳天祥,张文江,余勇. 基于神经网络模型的中国表层土壤有机质空间分布模拟方法[J]. 地球科学进展, 2012, 27(2): 175-184.
[13] 朱子先, 臧淑英. 基于遗传神经网络的克钦湖叶绿素反演研究[J]. 地球科学进展, 2012, 27(2): 202-208.
[14] 崔景伟,邹才能,朱如凯,白斌,吴松涛,王拓. 页岩孔隙研究新进展[J]. 地球科学进展, 2012, 27(12): 1319-1325.
[15] 刘永和, 郭维栋, 冯锦明, 张可欣. 气象资料的统计降尺度方法综述[J]. 地球科学进展, 2011, 26(8): 837-847.