Land Cover Classification Using ALOS Image Based on Textural Features and Support Vector Machine

Abstract

The high spatial resolution remote sensing images are used widely in the land cover classification;nevertheless,the traditional pixel-based classification has the weakness of relatively low accuracy.For the purpose of improving the accuracy of the high spatial resolution image classification,the textural features were extracted quickly by using the method of Gray Level Co-occurrence Matrices(GLCM),and then the ALOS image of the typical test area in Huzhou city of Zhejiang province was classified based on textural features and Support Vector Machine(SVM).The results show that image classification based on textural features and SVM can better extract surface features with precision of 90.88%.The classification precision based on SVM only is higher than that based on maximum likelihood,with the former precision being 89.96% and the latter 86.16%.Extracting land cover types quickly and accurately can provide a service for the research on appearance and spatial-temporal distribution of the agricultural non-point pollution source,and also provide scientific evidence for exploration of reasonable land use model and sustainable land utilization in Taihu basin.