In order to explore the application of China-Brazil Earth Resources Satellite (CBERS) remote sensing data to urban land cover/land use analysis，the authors developed the decision tree classifier，whose generation strategy is discussed in detail in this paper. With Xuzhou city as the study area，five features，i.e.，near-infrared band，Global Environment Monitoring Index (GEMI)，NDVI，and the first and second principal components，were extracted and used for urban land use classification. On the basis of experiments，the decision tree was designed based on prior knowledge and statistical analysis，and a new interactive decision tree generation strategy was developed to optimize threshold selection. A comparison of the classification results with results of Maximum Likelihood Classifier (MLC) and Support Vector Machine (SVM) classifier shows that the decision tree classifier that uses multiple features is effective in land use classification from CBERS imagery.