Gaussian Mixture Models for Supervised Classification of Remote Sensing Multispectral Images

摘要

This paper proposes the use of Gaussian Mixture Models as a supervised classifier for remote sensing multispectral images. The main advantage of this approach is provide more adequated adjust to several statistical distributions, including non-symmetrical statistical distributions. We present some results of this method application over a real image of an area of Tapajos River in Brazil and the results are analysed according to a reference image. We perform also a comparison with Maximum Likelihood classifier. The Gaussian Mixture classifier obtained best adjust about image data and best classification performance too.