Land use classification based on support vector machine in karst areas

The classification of land use in karst areas is mainly through the interpretation of satellite images to get. The traditional interpretation methods are supervised classification and unsupervised classification. But the classification polygons is trivial by supervised classification, and boundary is also complex. Different categories can be distincted by unsupervised classification, however, the property can´t be determined by it. SVM(support vector machine) is a new type image classification technique, it has advantages of high accuracy, few errors and misclassifications. In this study, we use SPOT image data, topographic maps and administrative divisions data, employ the knowledge of support vector machine and land use classification, on the support of ENVI software, classify land use of the study area by SVM classification, supervised classification and unsupervised classification. The results showed that using the SVM to classify land use, the accuracy is high, while the supervised classification´s and unsupervised classification´s are low.