Using Consolidated Covariance Image for Discrimination of Habitats

Abstract

In this paper the method for transforming the multispectral image is defined, based on the Cholesky decomposition of empirical covariance matrices of pixels within a chosen window and consecutive calculation of mean values of the triangular matrix elements. This procedure is called the covariance consolidation method and it is applied to three subsets of the spectral bands thus transforming p-dimensional image into the 3 dimensional Consolidated Covariance Image (CCIm). CCIm is proposed to visualize the spectral diversity of remotely sensed objects. Within the described study, CCIm was created from the 15-band multispectral image to perform visual analysis of the nature park "Dviete floodplain" in Latvia. It was shown that CCIm provides complementary information about the habitats that can be used for their discrimination. CCIm can be used together with Principal Component Analysis (PCA) or other methods to classify regions of interest.

Simplified Classification of Multispectral Image Fragments

“…It is seen that hogweed areas are speckled and characterized by a large variance of pixel values in visualized spectral bands. Therefore initially application of the Consolidated Covariance Image [10] (CCIm) to detection of hogweed areas was considered. It allows the image to be transformed to the form where the increased covariance around a pixel results in brighter colours of that pixel.…”