Combinational approach of object oriented land cover image classification

Applied technique of object-based land cover image classification for support vector machines were studied. And a combinational approach was estiblished, namely FG-LSSVM, with least squares support vector machines (LSSVM) and fuzzy and grey degree of correlation (FG), which was a feasible high-precision image classification algorithm for land cover. According to the spatial scale and spectral characteristics of different targets on rectified image, the number of objects was automatically determined by using the stable gradient of kernel density algorithm, in which local objects with unique identifier in arbitrary shapes were picked up. To compare the performance of the presented method with that of other object oriented methods, with original samples, three models were successively verified, which were standard support vector machines (SVM) and the fuzzy nearness improved support vector machines (FSVM), and the traditional K nearest neighbor (KNN) object-oriented methods. A high precision land cover image classification system was established with the proposed approach. The results showed the total precision of FG-LSSVM was about 2.4% higher than that of SVM, FSVM and KNN object-oriented methods in the study area. The proposed method also meets the requirements of land cover image classification in respect of efficiency and effects.