Focusing on the issue that current real-time map matching algorithms are difficult to keep high efficiency and high accuracy simultaneously, an improved dynamic weighted real-time map matching algorithm was proposed. Firstly, considering the temporal, speed, heading and direction constraints of Global Positioning System (GPS) points and the topological structures of road network, a weighted model was constructed in the algorithm based on spatio-temporal analysis, which consisted of proximity weight, heading weight, direction weight and connectivity weight. Then according to the properties of GPS points, a dynamic weighted coefficient model was created. Lastly, the best matching road segment was selected according to the confidence level of current GPS point. The experiments were conducted on three city bus trajectories with length of 36 km in Chongqing. The average matching accuracy of the algorithm was 97.31% and the average matching delay of each GPS point was 17.9 ms. The experimental results show that compared with the contrast algorithms, the proposed algorithm has higher accuracy and efficiency, and has better performance in matching Y-junctions and parallel roads.