Each city has different spatial features of urban development. These features are often affected by local urban development strategies and planning policies at that time. As the capital and most populous city in China, the urban built-up area of Beijing has been enjoying a continuing expansion during its rapid urbanization in the last twenty years. There are some specific rules and spatial features in the urban land construction during Beijing’s urban expansion. In this paper we present a non-negative matrix factorization (NMF) method and recognize these patterns by using the land construction permission dataset. We find out two main developing patterns in these years, and made a series of analysis and explanation on both patterns by combining some additional information. This pattern discovery method could be easily extended to other area of urban studies, and the results would be helpful for the urban designers and policy makers.