Abstract: In order to investigate the influence of built environment and other factors on the severity of pedestrian accidents, a random parameters Logit model with heterogeneity in means and variances was established to analyze the heterogeneity of pedestrian accident injury severities at intersections. By use of the data of pedestrian collision accidents, 27 influencing factors were selected from six aspects such as pedestrians and drivers, vehicles, roads, environment, time and built environment. The elasticity coefficient was used to analyze the influence of significant variables on accident injury severities. The results show that the existing of the accessibility of subway station is a random parameter, and setting the vehicle status as other will reduce the mean value of the random parameter and pedestrian accident injury severities at the intersections. Variance is significantly related to afternoon, which will increase the dispersion of random parameter, widen the distribution and increase the randomness. The impact of different land use types and intersection types on pedestrian accident injury severities is significantly different. When the accessibility of commercial areas and subway stations exists, the probability and severity of pedestrian accidents at intersections decrease. Moreover, factors such as pedestrian misconduct significantly increase the severity of pedestrian accidents at intersections. The research results provide a reference and basis for formulating the optimization and improvement measures of the built environment of intersections.