Authors: Yingbin Deng*, University of Wisconsin Milwaukee
Topics: Remote Sensing, Land Use and Land Cover Change, Quantitative Methods
Keywords: spectral transformation, spectral mixture analysis, spectral band normalization
Session Type: Paper
Start / End Time: 8:00 AM / 9:40 AM
Room: Lafayette, Marriott, River Tower Elevators, 41st Floor
Presentation File: No File Uploaded
Although a number of studies applied spectral transformations to spectral mixture analysis (SMA), few of them discussed the necessity and applicability. Therefore, this paper focus on two questions: 1) whether significantly different results would be generated through applying a spectral transformation, and 2) which spectral transformation performs better in urban areas. In particular, we examined twenty-one spectrally transformed schemes as well as no-transformed scheme in three different study areas (Janesville, WI; Ashville, NC; and Columbus, OH). The transformed schemes include first-derivative, second-derivative, third-derivative, independent component analysis (ICA), minimum noise fraction transform (MNF), principle component analysis (PCA), tasseled-cap transformation (TC), continuum removal (CR), Gaussian high pass (GHP), Gaussian low pass (GLP), high pass (HP), low pass (LP), normalization by band (Norm), normalization by pixel (Norm_wu), tie transformation from band 1 to 6 (Tie1-7). Each transformed scheme was tested 100 times using different endmember classes’ spectra. Paired-sample T test was employed to test the significance of mean difference between each transformed scheme and no-transformed scheme. Results illustrated that all spectrally transformed schemes except Norm_wu did not improve SMA’s accuracy significantly in three study areas. We concluded that we can apply spectral band normalization (Norm_wu) to acquire more accurate SMA results in urban environment.