Monitoring urban spatial information is vital to reveal the relationship between the human activity and environment, especially in the Chinese Silk Road Economic Belt, so as to allocate resources reasonably and realize sustainable development. To promote the remote sensing application in this field, a new method was proposed for urban built-up areas extraction mainly based on the support vector machine (SVM) classification with iterative sample refinement, combining Defense Meteorological Satellite Program-Operational Linescan System (DMSP-OLS) nighttime light data, and other auxiliary data such as Landsat images and the GlobeLand30 land cover product. Experiments were conducted by using the proposed approach for several cities in the southwest of the Chinese Silk Road Economic Belt, as classified by statistics and Landsat images. Compared with the traditional threshold dichotomy method and the state-of-the-art improved neighborhood focal statistics (NFS) method, the proposed method achieved better performance with respect to less relative error, and higher overall accuracy and Kappa coefficient.