Density peak clustering can be used in identifying high-density regions for urban hot spots detection. The distance matrix of each two position points needs to be calculated in the existing density peak clustering methods which causes inefficient clustering when processing large-scale data, and the traditional two-dimensional decision map cannot identify the coincident points. Thus, characteristic density peak clustering algorithm is proposed to avoid the influence of noise. At first, the location feature points and support index are defined to simulate the original locations. The number of feature points is adjusted by parameters to make density peak clustering no longer sensitive to the amount of data to simplify the complexity to be solved. And then, the local density and the distance between the clustering centers of the feature points are proposed to construct three-dimensional decision map. Finally, the clustering center, basic clustering points, and noise data points are determined using the three-dimensional decision map combined with the support index of the feature points. Experiments are performed on real data set and the prototype system to verify that the method can significantly improve time efficiency while clustering accuracy is maintained.