Dominant attitudes of structural plane in rock mass are fundamental to the research and analysis of geotechnical engineering. Conventional rose diagram and pole isodense diagram are of poor accuracy, just with results of the number of dominant groups. In order to accurately obtain the dominant attitudes of structural plane, we proposed an improved fast clustering analysis method to turn structural plane attitudes into points on hemisphere surface of the unit space. According to structure density functions, we calculated the weighted values of every data point and classified the attitudes of structural planes. Furthermore, we applied the method to the statistical analysis of dominant structural plane of a slope near a mine transportation road. Results show that this method is reliable and reasonable. We can use it to determine the dominant attitudes of structural plane, and effectively avoid adverse effect of isolated points on the clustering result.