A Classification Algorithm for Hyperspectral Images Based on Synergetics Theory

Abstract

This paper presents a classification methodology for hyperspectral data based on synergetics theory. Pattern recognition algorithms based on synergetics have been applied to images in the spatial domain with limited success in the past, given their dependence on the rotation, shifting, and scaling of the images. These drawbacks can be discarded if such methods are applied to data acquired by a hyperspectral sensor in the spectral domain, as each single spectrum, related to an image element in the hyperspectral scene, can be analyzed independently. The spectrum is first projected in a space spanned by a set of user-defined prototype vectors, which belong to some classes of interest, and then attracted by a final state associated to a prototype. The spectrum can thus be classified, establishing a first attempt at performing a pixel-wise image classification using notions derived from synergetics. As typical synergetics-based systems have the drawback of a rigid training step, we introduce a new procedure which allows the selection of a training area for each class of interest, used to weight the prototype vectors through attention parameters and to produce a more accurate classification map through plurality vote of independent classifications. As each classification is in principle obtained on the basis of a single training sample per class, the proposed technique could be particularly effective in tasks where only a small training data set is available. The results presented are promising and often outperform state-of-the-art classification methodologies, both general and specific to hyperspectral data.

Noise Reduction in Hyperspectral Images Through Spectral Unmixing

“…A recently described classification methodology for hyperspectral data based on synergetics theory [14] projects any image element onto a "semantic" subspace, in which every dimension represents the similarity to a given class of interest. This procedure inspires a supervised methodology based on spectral unmixing to suppress noise for the bands with low SNR described in the previous section, but that turns out to be effective on any band of a given hyperspectral image.…”

“…Ground truth is available for 15 classes in the area reported in Fig. 3(f), and we select the average spectrum over a 6 × 6 pixels area for each class (the size of the averaging window has been set empirically), resulting in 16 reference spectra (the class corn has been divided in two classes as in [14]). The next step is the choice of the unmixing algorithm for Eq.…”

Fast TLS Denoising Algorithm Using Grid Technique

“…Based on the use of image patches, Hirakawa [1] proposes a Total Least Squares (TLS) denoising algorithm. This method allows us to take into account the system uncertainties and uses a linear combination of noisy image patches and total least squares technique [16], [17] to remove noise. In spite of having a certain denoising performance on mixed noise, it has a very low calculation efficiency.…”