Position-Aware Non-negative Matrix Factorization for Satellite Image Representation

Abstract

Satellite images clustering is a challenging problem in remote sensing and machine vision, where each image content is represented by a high-dimensional feature vector. However, the feature vectors might not be appropriate to express the semantic content of images, which eventually leads to poor results in clustering and classification. To tackle this problem, we propose a novel approach to generate compact and informative features from image content. To this end, we utilize geometrical information (as meta data accompanied with images) in the context of Non-negative Matrix Factorization (NMF) to generate new features. We assess the quality of new features by applying k-means clustering on the generated features and compare the obtained clustering results with those achieved by original features. We perform experiments on several satellite image data sets represented by different state-of-the-art features and demonstrate the effectiveness of the proposed method.