As the number of new energy vehicles equipped with autonomous driving technology increases, the degree of interaction between vehicles and pedestrians is different, which affects the accuracy of pedestrians’ judgment of vehicle behavior and pedestrians’ behavior of crossing the street. Therefore, it is necessary to study road traffic safety. In this paper, SP questionnaire was used to obtain data of pedestrians’ intention to take risks across the street. Based on the theory of extended planned behavior, the formation mechanism model of pedestrians’ intention to take risks across the street was established. On the basis of traditional dimensions, two variables of trust and risk perception were expanded, and a structural equation model was used to verify and improve the model. From the aspect of path relationship and the effect of each factor, analyze the relationship between the variables and the degree of influence. The results show that the extended TPB model can effectively explain pedestrian risk-taking behavior, and the role of risk perception, trust,and perceived behavior control on risk-taking behavior intention is more prominent. The k-means clustering method is adopted to cluster the people who take risks to cross the street, which can be divided into ignoring rules (risk-taking type), flexible following rules (stable type),and strict following rules (prudent type). Therefore, the study of pedestrian risk behavior intention is helpful to traffic safety management and improves traffic efficiency.