This paper developed an approach for urban land use mapping with QuickBird stereo imageries. The accuracy of ground point determination was first improved by RPCs regeneration using the second-order polynomial model for bias correction in the image space. In urban land use mapping, DEM was generated from the stereo pair using the refined RPCs and then an orthoimage was generated. Urban features such as roads, greenbelts, and waters were digitized from the orthoimage, and buildings were extracted using an approach of stereo positioning. The result demonstrated that a planimetric mapping accuracy of 0.8 m could be achieved.