Semantic Supervised Clustering to Land Classification in Geo-Images

Abstract

In this paper, we propose a semantic supervised clustering approach to classify lands in geo-images. We use the Maximum Likelihood Method to generate the clustering. In addition, we complement the analysis applying spatial semantics to improve the classification. The approach considers the a priori knowledge of the multispectral image to define the training sites (classes) related to the geographic environment. In this case the spatial semantics is defined by the spatial properties, functions and relations that involve the geo-image. By using these characteristics, it is possible to determine the training data sites with a priori knowledge. This method attempts to improve the supervised clustering, adding the intrinsic semantics of the geo-images to determine the training sites that involve the analysis with more precision.