In order to reduce potential safety hazards caused by passenger agglomeration at passenger stations during peak periods, based on full sample ticket purchasing data of highway passenger transport, a quantitative model of agglomeration behavior was established from three aspects, which were agglomeration probability, duration and intensity, and gathering characteristics of passenger flows at different levels of stations during weekdays, holidays and Spring Festival were obtained. Then, considering passengers' individual attributes and travel characteristics, agglomeration behavior characteristics of different groups were analyzed. Finally, a difference test was conducted to analyze influencing factors of such behavior, and countermeasures to reduce security risks were put forward. The results show that probability of passenger agglomeration in highway stations follows distribution of six polynomials, and its cumulative probability follows that of quadratic function, with highest probability of it occurring [10,20]min before departure. Regardless of travel distance, most travelers have a short waiting time, but in different periods, their flow concentration intensity and speed for different levels of passenger stations are significantly different. The spatial and temporal clustering characteristics of the first level stations are significant, while randomness of second level ones is stronger.