The paper proposes a novel method for integrating urban POI and road networks based on semantic association. The proposed method utilizes semantic information involved in each POI point and road segment to construct a POI semantic connectivity graph which represents the similar structural patterns with road network. Then the paper recognizes and simplifies road junctions so as to record the correct position and connected road names of each junction. After that, the matching feature points between POI and road data can be found by comparing the connected road names of POI and road junction. At last, an iterative process is implemented to geometrically integrate two heterogeneous data sets of POI and road network. The experimental results in Wuhan show that our method can effectively match POI points to road networks, and ensure the relatively positional relations between POI points and the neighboring roads.