DocumentCode :
124663
Title :
Two inverse processes: Spectral reconstruction and pixel unmixing
Author :
Lei Yan ; Sui-hua Liu ; Hui-li Liu ; Xin Jing ; Chengqi Cheng ; Hong Wang
Author_Institution :
Inst. of Remote Sensing & Geographic Inf. Syst., Peking Univ., Beijing, China
fYear :
2014
fDate :
11-14 June 2014
Firstpage :
462
Lastpage :
469
Abstract :
The application of hyperspectral remote sensing has been a research focus in recent years, and one of its fundamental goals is to detect and classify the constituent materials for each pixel in the scene, which is pixel unmixing. This research proposes spectral reconstruction, which is the inverse process of pixel unmixing. It can be used to analyze the essence of the hyperspectral imaging procedure and provide inverse analysis to the pixel unmixing. Also, it may find the physical origin of hyperspectral sensors´ parameters degenerating with the stability of multispectral scanner, which helps to provide quantitative basis for the development and improvement of hyperspectral sensors. Based on research for years, we can get hyperspectral data of high quality and stability from multispectral data, which is a new way to the hyperspectral application. The technical route is using the normalized multiple endmember decomposition method (NMEDM) to decompose the endmember data of vegetation, water and soil based on the condition of fuzzy sets and full constraint. The characteristics are as follows: this new method considers the space-time variation of the pixel endmember, the hyperspectral reconstruction can be achieved with less calculated amount, and the terrain spectral reflectance can be showed with less data. This research includes: deeper mining of the multispectral information, sensitivity analysis of multispectral bands, verification of the hyperspectral reconstruction model based on the ground multispectral imaging.
Keywords :
fuzzy set theory; geophysical image processing; hyperspectral imaging; image reconstruction; inverse problems; remote sensing; sensitivity analysis; vegetation; NMEDM; deeper mining; full constraint; fuzzy sets; ground multispectral imaging; hyperspectral imaging procedure; hyperspectral reconstruction model; hyperspectral remote sensing; hyperspectral sensor parameters; inverse analysis; inverse processes; multispectral bands; multispectral information; multispectral scanner; normalized multiple endmember decomposition method; pixel endmember; pixel unmixing; quantitative basis; sensitivity analysis; soil; space-time variation; terrain spectral reflectance; vegetation; water; Earth; Educational institutions; Hyperspectral imaging; Image reconstruction; Reflectivity; fuzzy sets; hyperspectral remote sensing; multiple endmember; pixel unmixing; spectral reconstruction;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Earth Observation and Remote Sensing Applications (EORSA), 2014 3rd International Workshop on
Conference_Location :
Changsha
Print_ISBN :
978-1-4799-5757-6
Type :
conf
DOI :
10.1109/EORSA.2014.6927934
Filename :
6927934
Link To Document :
بازگشت