Naturalistic driving recorders were installed on 11 passenger cars, running in 5 Chinese cities ranging from first tier to third tier cities to obtain naturalistic driving data. And 65 near-crash cases related to pedestrians are extracted from the database and researched. Based on vehicle’s speed obtained from OBD (On-Board Diagnostic), image process method and kinematic formulas, information of the pedestrian, road environment and vehicle is collected. Firstly, based on the 65 samples, qualitative analysis on the key elements such as pedestrian’s walking direction and road congestion status is conducted to obtain characteristics of near-crash cases related to pedestrians in China. Secondly, typical scenarios at the time of risk start (TRS) are obtained from 39 samples through cluster analysis. Thirdly, the differences between the current study and previous studies are analyzed and discussed further.