A new automatic approach based on multiple features integration is proposed to interpret urban remote sensing images more effectively and comprehensively.The approach builds a hierarchical objects network to organize image structure and gets precise processing units.Then the probabilistic learning integrating multiple features including colour,texture,shape and position are performed to train a best classifier,and label all of the objects according to their classification values.The approach also applies spatial smoothing which incorporates contextual information to eliminate the adverse effects caused by background disturbance,occlusion and so on.After vectorization procedure,final results are given.Experiments demonstrate that proposed approach achieve high exactness and robustness in interpreting manifold urban remote sensing images.