Victor-Emil Neagoe, Vlad Berbentea Chirila
unsupervised classification, hyperspectral imagery, Earth Observation (EO), Gaussian mixture model (GMM), expectation-maximization (EM)
This paper presents an approach for improving performances of the unsupervised classification by proposing a technique to combine the classical techniques of K-means clustering and the Gaussian mixture model (GMM) based on expectation-maximization (EM). The proposed model means to apply firstly K-means clustering, and to use the result of this technique to compute means µk, covariance matrices Sk and mixing coefficients pk, considered as initialization parameters for the next stage of GMM-EM. The above mentioned algorithm has been successfully applied for clustering of Earth Observation (EO) hyperspectral imagery. The information content of hyperspectral images with hundreds of channels allows us to remotely identify ground materials, based on their spectral signature. The performances of the proposed method have been evaluated using the Pavia Centre hyperspectral database with 102 spectral bands and a resolution of 1.3 meters/pixel. The proposed combined clustering model K-means+GMM-EM, has led to a significant improvement in performance over any of the two single clustering techniques K-means and GMM-EMM.
Cite this paper
Victor-Emil Neagoe, Vlad Berbentea Chirila. (2016) A New Approach to Unsupervised Classification of Hyperspectral Earth Observation Imagery Using a Gaussian Mixture Model. Signal Processing, 1, 134-137