遙感數(shù)據(jù)的貝葉斯網(wǎng)絡(luò)分類研究
The Study on Remote Sensing Data Classification Using Bayesian Network
-
摘要: 由于遙感成像過(guò)程的復(fù)雜性，遙感數(shù)據(jù)中包含了一定程度的不確定性因素。利用最大似然分類器處理遙感數(shù)據(jù)時(shí)分類精度受一定的影響, 為了提高分類精度往往需要引入先驗(yàn)知識(shí)。貝葉斯網(wǎng)絡(luò)是一個(gè)帶有概率注釋的有向無(wú)環(huán)圖，可以動(dòng)態(tài)地對(duì)先驗(yàn)概率密度修正，提高分類精度，也沒(méi)有嚴(yán)格的數(shù)據(jù)正態(tài)分布前提要求，適合處理不完整復(fù)雜的數(shù)據(jù)。該文介紹了利用貝葉斯網(wǎng)絡(luò)對(duì)遙感數(shù)據(jù)進(jìn)行分類處理的算法和技術(shù)過(guò)程。分類結(jié)果表明：貝葉斯網(wǎng)絡(luò)具有穩(wěn)定的數(shù)學(xué)基礎(chǔ)，是一種可供遙感信息處理領(lǐng)域利用的有效新方法。
-
關(guān)鍵詞:
- 遙感數(shù)據(jù);貝葉斯網(wǎng)絡(luò);分類
Abstract: Because of the complexity in satellite remote sensing imaging system, some uncertainties or mixed spectrum information are contained in the data. By using maximal likelihood classification to process remote sensing data, the result accuracy of the classification is affected. In order to improve the accuracy of the classification, prior knowledge is needed to modify the probability. Bayesian network is composed of directed acyclic graph and probability chart; it can modify the prior probability density dynamically and improve the accuracy of classification. In this paper, a technical procedure is demonstrated that using Bayesian network to process the remote sensing data, the classification results prove that Bayesian network has solid mathematics base and can be a new effective methods for remote sensing data processing. -
Richard O D, Peter E H, David G S著,. 李宏?yáng)|,姚天翔等譯, 模式分類. 北京：北京機(jī)械工業(yè)出版社, 2001: 16-81.[2]周顏軍，王雙成，王輝. 基于貝葉斯網(wǎng)絡(luò)的分類器研究. 東北師大學(xué)報(bào)自然科學(xué)版, 2003, 35(2): 21-27.[3]Bob Liang. 未來(lái)處理器架構(gòu)進(jìn)行應(yīng)用驅(qū)動(dòng)的研究報(bào)告,Intel Microprocessor Research 2002 Forum, Beijing China , October 29, 2002: 1-20.[4]Cheng, J, Greiner R, Kelly J, Bell D A, Liu W. Learning Bayesian networks from data: an information-theory based approach[J].The Artificial Intelligence Journal.2002, 137(1):43-90[5]Hurn M A, MardiaKV. Bayesian fused classification of medicalImages[J].IEEE Trans. on Geoscience and Remote Sensing.1999, 37:1292-1305[6]哈斯巴干, 馬建文, 李啟青, 韓秀珍, 劉志麗. 基于小波變換的ASTER數(shù)據(jù)的自組織特征映射神經(jīng)網(wǎng)絡(luò)分類研究, 中國(guó)科學(xué)(D輯), 2003, 33(9): 896-902.[7]Ma Jianwen, Guo Huadong, Wang Changlin, et al.. Extraction of polymetallic mineralization in formation from multi-spectral thematic mapper data using the Gram-Schmidt orthogonal projection (GSOP) method[J].Int. J.Remote Sensing.2001, 22 (17):3323-3337[8]李啟清，馬建文，哈斯巴干等. 基于貝葉斯網(wǎng)絡(luò)模型的遙感數(shù)據(jù)處理技術(shù). 電子與信息學(xué)報(bào), 2003, 25(10): 132-136.[9]范明, 孟小峰等. 數(shù)據(jù)挖掘概念與技術(shù). 北京：機(jī)械工業(yè)出版 社, 2001: 196-200.[10]Cooper G F, Herskovtis E. A Bayesian method for the induction of probabilistic network from data. Machine Learning. 1992, 9(4): 309-347.[11]史忠植. 知識(shí)發(fā)現(xiàn). 北京：清華大學(xué)出版社, 2002: 169-198. -
計(jì)量
- 文章訪問(wèn)數(shù): 2412
- HTML全文瀏覽量: 118
- PDF下載量: 1524
- 被引次數(shù): 0
下載:
