Hyperspectral ratio feature selection: agricultural product inspection example

Abstract

We describe a fast method for dimensionality reduction and feature selection of ratio features for classification in hyperspectral data. The case study chosen is to discriminate internally damaged almond nuts from normal ones. For this case study, we find that using the ratios of the responses in several wavebands provides better features than a subset of waveband responses. We find that use of the Euclidean Minimum Distance metric gives slightly better results than the more conventional Spectral Angle Mapper distance metric in a nearest neighbor classifier.

Hyperspectral Image Processing Methods

“…Although the "feature selection" terminology has been widely used and recognized in many fields, this feature selection problem is also commonly known as a band selection problem in the hyperspectral imaging literature (Bajwa et al 2004;Nakariyakul and Casasent 2004;Martínez-Us o et al 2006;Su et al 2008). In fact, a band selection problem is a particular case of feature selection, specifically to reduce the dimensionality of hyperspectral images and find the important and useful features, i.e., wavelengths, for analysis, classification and regression.…”

Wavelet-SOM in feature extraction of hyperspectral data for classification of nematode species

“…In [4], authors applied DWT to the hyperspectral signatures, and used the coefficients as features for classification purposes. For this study, authors combine the idea utilized in [1] and [2] to extract important but reduced set of features. The idea of combining wavelets with neural network can also be seen in [16] and [17].…”

“…Hence it is vital to reduce the dimensionality of the data sample to an acceptable lowdimensional feature vector without losing vital information required for analysis [1]. Another reason to reduce the dimension is to avoid the problem of "curse of dimensionality" [2]. It is also seen that remotely sensed data is highly correlated.…”