VHR satellite image segmentation based on topological unsupervised learning

摘要: High spatial resolution satellite imagery has become an important source of information for geospatial applications. Automatic segmentation high-resolution is useful obtaining more timely and accurate information. In this paper we introduce a new approach automatic image into different regions (corresponding to various features texture, intensity, color) based on topological un-supervised learning. Three types methods were studied in work: matrix factorization, self-organizing maps probabilistic models. The approaches applied real Very Resolution (VHR) the French city Strasbourg. obtained results validated using internal external clustering validation indexes.