The performance of pixel window algorithms in the classification of habitats using VHSR imagery

摘要: This study investigates the potential of three advanced pixel window classification methods for habitat mapping, namely Kernel based spatial Re-Classification (KRC), Radial Basis Function (RBF) neural networks (NN) and Support Vector Machines (SVM). KRC classifier takes into account arrangement frequency spectral classes present within a predefined square kernel. On other hand, RBF-NN SVM classifiers use set parameters (digital numbers training pixels, mean values standard deviations specified kernel) as input information. The fuzzy means clustering algorithm is utilized RBF networks. method on partition space requires only short amount time to determine both structure classifier. radial basis function also adopted kernel in implementation methodology. test area Lake Kerkini, wetland ecosystem located Macedonia (Northern Greece). are applied very high resolution multispectral satellite image acquired by IKONOS-2. nomenclature used EUNIS, detailed hierarchical scheme. Several experiments carried out using same samples order behaviour perform meaningful comparisons. Overall, all performed satisfactorily; however consistently outperformed KRC, reaching overall accuracies 72% 69%, respectively.