To tackle the problems existent in road information extraction from high resolution remote sensing,the authors put forward an improved approach to road extraction based on watershed segmentation according to the basic theories of object-oriented method and mathematical morphology.Firstly,the image is processed by improved watershed segmentation to extract basic road information after preprocessing.Then object-oriented method is used to extract road per-parcel so as to optimize the road extraction results.Finally,after binary image processing,the incomplete results can be removed and corrected by using mathematical morphological transformation.Experimentation shows that the proposed method can extract urban road information efficiently and process the roads from the complex urban context fairly satisfactorily.