Abstract: To scientifically quantify the impact of bowed pedestrians on traffic efficiency in rail transit channels, this paper, based on the field collection of video data of bowed pedestrians in rail transit channels, extracted the bowing frequency and raising frequency of pedestrians, walking speed and turning angle and other parameters by using video automatic recognition technology. The compensatory behavior characteristics of the bowed pedestrians were analyzed from the strategic and operational levels, and the similarities and differences between the normal pedestrians and the bowed pedestrians in the corners, walking time, and walking speed were verified. Using the cellular automata model based on the dynamic parameter model, combined with the compensating behavior characteristics of the bowed pedestrians, a numerical simulation model of the bowed pedestrians was constructed, and the density, travel time, individual space occupied area and the flow rate of the pedestrians with different mixing rates were analyzed. Results show that more than 30% of pedestrians with bowed heads have a significant impact on pedestrian flow. This study can help analyze the characteristics of pedestrian behavior in depth, improve the application of simulation models in simulating multiple types of complex pedestrian flows, and provide a theoretical basis for the planning of rail transit internal facilities and passenger flow control.