There is congestion in urban rail transit carriage, which directly exerts an effect on the comfortability of passengers and operational efficiency of unban transportation network. Based on different physical and psychological requirements of passengers and some calculation on passengers’ rate of mixture in urban rail transit carriage, with the investigation results of passengers’ choice behavior of standing position, age, gender and other indicators, density of standing passengers evaluation criteria is established based on calculation of passengers mixed degree. To accurately identify the number of passengers, gender and age in the key points and regions of carriage, the paper choose the method of Regional probability estimation and deep learning. According to the output of model, it can be judged whether or not the carriage is congested. Meanwhile, the type of congestion may be determined. The method can rapidly identify congestion of carriage situation and determine whether the type of carriage congestion belongs to frequent or disequilibrium congestion.

CICTP