With the rapid development of urban traffic, accurate and up-to-date road maps become a crucial demand for human’s daily life and urban traffic controlling. Recently, with the emerging of crowdsourcing mapping, surge attention is paid on generating road networks from spatio-temporal trajectory data. However, existing methods rarely explore the road changing patterns contained in multi-temporal trajectory data and difficultly satisfy the demand of precision and efficiency. Hence, the paper proposed a hybrid method for incrementally extracting urban road networks from trajectory data. Firstly, the single-temporal road networks are initialized by mathematical morphology method. Multi-temporal road networks are then adjusted based on a gravitation force model. Finally, road networks are incrementally constructed and geometrically delineated using k-segment fitting algorithm. Several case studies are conducted to demonstrate our method can effectively improve the efficiency and precision of road extraction and correctly mine the incremental change patterns of road networks from multi-temporal trajectory data to help with road map renewal.