本文選題：高光譜圖像 + 混合像元分解�。� 參考：《成都理工大學(xué)》2014年碩士論文

【摘要】：上世紀(jì)80年代高光譜遙感技術(shù)開(kāi)始發(fā)展，隨著成像技術(shù)不斷成熟，，高光譜遙感在越來(lái)越多領(lǐng)域得到廣泛應(yīng)用。高光譜圖像與傳統(tǒng)的多光譜遙感圖像相比，能夠獲得每個(gè)像元的連續(xù)波譜信息，這解決了“成像無(wú)光譜”，“光譜不成像”的技術(shù)難題。然而由于成像光譜儀分辨率不高等原因，這使得圖像中的每個(gè)像元內(nèi)往往包含著多種不同的地物類(lèi)型，從而形成混合像元，在信息提取過(guò)程中不能隨便將這些像元?jiǎng)潥w到任何一類(lèi)地物中去。若把只包含一種純凈地物的像元稱為端元，把端元從高光譜圖像中準(zhǔn)確的提取出來(lái)成了研究的重點(diǎn)。獲取端元后，才能很好地進(jìn)行后續(xù)的解混、匹配、分類(lèi)識(shí)別等研究。近年來(lái)國(guó)內(nèi)外學(xué)者研究和發(fā)展了不少混合像元分解模型，其中線性混合模型以其結(jié)構(gòu)簡(jiǎn)單，物理意義明確等優(yōu)點(diǎn)成為研究的熱點(diǎn)。 本文介紹了高光譜圖像的數(shù)據(jù)特點(diǎn)，高光譜圖像的降維，端元提取的經(jīng)典算法，像元的線性混合模型等基礎(chǔ)理論。其中在數(shù)據(jù)降維方面主要采用了主成分分析法和最小噪聲分離法，并采用了純凈像元指數(shù)和內(nèi)部體積最大法進(jìn)行了端元提取。然后研究了線性模型下的線性波譜分和基于MTMF的混合像元分解，其中線性波譜分離主要采用全約束下最小二乘法來(lái)求豐度。最后重點(diǎn)介紹了MTMF方法的理論基礎(chǔ)和處理流程，即結(jié)合MNF變換，使用匹配濾波MF進(jìn)行豐度估計(jì)，再用MT檢查并排除假陽(yáng)性值。模擬數(shù)據(jù)實(shí)驗(yàn)結(jié)果表明MTMF方法誤差較小，同時(shí)，使用MTMF法對(duì)ENVI自帶的高光譜圖像數(shù)據(jù)實(shí)驗(yàn)表明該方法有良好效果。最后，將MTMF法應(yīng)用到機(jī)載CASI/SASI高光譜遙感測(cè)量?jī)x器在吉木薩爾地區(qū)獲取的飛行測(cè)量高光譜SASI數(shù)據(jù)，對(duì)所識(shí)別出的端元進(jìn)行基于MTMF法混合像元分解，并進(jìn)行蝕變填圖，結(jié)果顯示與Google Earth所查找到的識(shí)別填圖區(qū)域所在的地質(zhì)風(fēng)貌基本吻合。
[Abstract]:Hyperspectral remote sensing technology began to develop in 1980s. With the development of imaging technology, hyperspectral remote sensing has been widely used in more and more fields. Compared with traditional multispectral remote sensing images, hyperspectral images can obtain the continuous spectral information of each pixel, which solves the technical problem of "imaging without spectrum" and "spectrum without image". However, because of the low resolution of the imaging spectrometer, each pixel in the image often contains many different types of ground objects, thus forming a mixed pixel. In the process of information extraction, these pixels can not be randomly classified into any kind of feature. If the pixel containing only one kind of pure ground object is called endelement, it is the focus of the research to extract the endmember from hyperspectral image accurately. After obtaining the endelements, we can do the following research, such as demultiplexing, matching, classification recognition and so on. In recent years, many mixed pixel decomposition models have been studied and developed by scholars at home and abroad. Among them, the linear mixed model has become a hotspot for its simple structure and clear physical meaning. This paper introduces the data characteristics of hyperspectral image, the dimension reduction of hyperspectral image, the classical algorithm of End-element extraction, the linear mixed model of pixel and so on. The principal component analysis (PCA) and the minimum noise separation (MNSS) are used to reduce the dimension of the data, and the pure pixel index and the maximum internal volume method are used to extract the endcomponents. Then the linear spectral fraction and the mixed pixel decomposition based on MTMF are studied in the linear model, in which the least square method with full constraints is used to calculate the abundance of the linear spectral separation. In the end, the theoretical foundation and processing flow of MTMF method are introduced emphatically, that is, combining with MNF transform, using matched filter MF to estimate abundance, then using MT to check and exclude false positive value. The simulation results show that the error of MTMF method is small, and the MTMF method has good effect on the hyperspectral image data of ENVI. Finally, the MTMF method is applied to the flight measurement hyperspectral SASI data obtained by airborne CASI/SASI hyperspectral remote sensing instrument in Jimusar area. The identified endelements are decomposed by mixed pixel method based on MTMF method, and altered mapping is carried out. The results show that the geological features of the identified mapping areas found by Google Earth are in good agreement with each other.
【學(xué)位授予單位】：成都理工大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類(lèi)號(hào)】：TP751

【參考文獻(xiàn)】

中國(guó)期刊全文數(shù)據(jù)庫(kù) 前10條

1 齊亞鳴;章劍龍;;基于遺傳神經(jīng)網(wǎng)絡(luò)算法的混合像元分解研究[J];安徽農(nóng)學(xué)通報(bào);2007年13期

2 趙銳;;混合像元線性分解的精度估算[J];測(cè)繪與空間地理信息;2010年06期

3 趙春暉;齊濱;王玉磊;;一種改進(jìn)的N-FINDR高光譜端元提取算法[J];電子與信息學(xué)報(bào);2012年02期

4 楊飛;張柏;劉志明;劉殿偉;王宗明;宋開(kāi)山;;玉米冠層FPAR的高光譜遙感估算研究——基于PCA方法及近、短波紅外波段[J];國(guó)土資源遙感;2008年04期

5 束炯;王強(qiáng);孫娟;;高光譜遙感的應(yīng)用研究[J];華東師范大學(xué)學(xué)報(bào)(自然科學(xué)版);2006年04期

6 蘇紅軍;杜培軍;盛業(yè)華;;高光譜遙感數(shù)據(jù)光譜特征提取算法與分類(lèi)研究[J];計(jì)算機(jī)應(yīng)用研究;2008年02期

7 陶秋香,陶華學(xué),張連蓬;線性混合光譜模型在植被高光譜遙感分類(lèi)中的應(yīng)用研究[J];勘察科學(xué)技術(shù);2004年01期

8 劉娟娟;王茂芝;葛世國(guó);李想;;全約束下高光譜混合像元線性分解[J];四川理工學(xué)院學(xué)報(bào)(自然科學(xué)版);2013年03期

9 吳波,張良培,李平湘;高光譜端元自動(dòng)提取的迭代分解方法[J];遙感學(xué)報(bào);2005年03期

10 呂長(zhǎng)春,王忠武,錢(qián)少猛;混合像元分解模型綜述[J];遙感信息;2003年03期

本文編號(hào)：1805376