Sub-pixel classification is a tough issue in remote sensing field. Although many kinds of software or its Module can be used to address this problem, their rationale, algorithms and methodologies are different, resulting in different use of different method for different purpose. This makes many users feel confused when they want to detect mixed feature content within a pixel and to use sub-pixel approach for practical application. It is necessary to make an in-depth comparison study for different sub-pixel methods in order for RS&GIS users to choose proper sub-pixel methods for their specific applications. After reviewing the basic theories and methods in dealing with sub-pixels, this paper made an introductory analysis to their principles, algorithms, parameters and computing process of three sub-pixel calculation methods, or Linear Unmixing in platform ILWIS3.0, Erdas8.5's Sub-pixel Classifier, eCognition3.0's Nearest Neighbor. A case study of three sub-pixel methods was then made of flood monitoring in Poyang Lake region of P.R.China with image data of band-1 and band-2 of NOAA AVHRR image. Finally, a theoretic, technological and practical comparison study was made of these three sub-pixel methods in aspects of the basic principles, the parameters to be set, the suitable application fields and their respective use limitation. Opinions and comments were presented in the end on the use of the sub-pixel calculation results of these three methods in a hope to provide some reference to future sub-pixel application study for the researchers in interest.
[1]
Giles M. Foody,et al.
Estimation of sub-pixel land cover composition in the presence of untrained classes
,
2000
.
[2]
P. Atkinson,et al.
Multiple-class land-cover mapping at the sub-pixel scale using a Hopfield neural network
,
2001
.
[3]
Peng Gong,et al.
An assessment of some factors influencing multispectral land-cover classification
,
1990
.
[4]
Hugh G. Lewis,et al.
Super-resolution target identification from remotely sensed images using a Hopfield neural network
,
2001,
IEEE Trans. Geosci. Remote. Sens..
[5]
Q. Zhan,et al.
Pixel unmixing at the sub-pixel scale based on land cover class probabilities : application to urban areas
,
2001
.
[6]
P. Fisher.
The pixel: A snare and a delusion
,
1997
.