Brushlet transform for hyperspectral feature extraction in automated detection of nutsedge presence in soybean

Wavelet-Based Feature Extraction for Improved Endmember Abundance Estimation in Linear Unmixing of Hyperspectral Signals

“…In this paper, a root mean square error (RMSE) of the endmember abundance estimation is defined for a quantitative evaluation. Given a set of testing data consisting of mixed-pixel spectra, the RMSE, , of abundance estimations for all the mixed-pixel spectra in the set is computed as (14) where represents the average error energy of the abundance estimation corresponding to the th mixed pixel in the set. In general, indicates an average deviation of the abundance estimate from the true abundance.…”

“…Moreover, the use of hyperspectal data causes another problem called the Hughes phenomena [10], which requires more training data for a supervised classification system to obtain accurate results. It has been realized that the use of feature extraction (or dimensionality reduction) can avoid the Hughes phenomena and improve the classification performance [10]- [14]. Naturally, the question is whether the linear unmixing performance can be improved by the use of appropriate features.…”