We process the Land Use/Cover Classification of the Shihezi Area by TM images from 2003 to 2013. With the rapid development of economy and urbanization in the region over the past decade the city has undergone rapid changes, the land use types and spatial structure over time, great changes have occurred. Based Remote Monitoring changes in Shihezi City land use and land cover, this dynamic of land use and land use data more accurate grasp, extended dynamic monitoring of urban construction land will be more scientific prediction trends changes in land use and land cover, land use constructive advice for decision-making departments for reference. However, due to the regional land cover type complex, difficult to distinguish between the image and is likely to cause misclassification. In this paper, support vector machine (SupportVector Machine, SVM) classification, by introducing a radial basis function nonlinear transformation mapping to high-dimensional space, extract them nonlinear characteristics, enhanced separability between different types, reduce misclassification phenomenon, to improve the accuracy of remote sensing image classification. Based on comparative analysis of accuracy obtained by SVM classification method is more in the traditional supervised classification method to improve the efficiency and accuracy of classification.