In real-world car to vulnerable road user (VRU) impact accidents, head impactisthe leading cause of VRU injuryanddeath. The currently used head impactor test used in pedestrian safety regulations/programs only considers the simple impact conditions that can not fully reflect the real impact event, and the used head injury criterion HIC onlyconsiders head liner acceleration. Meanwhile, specific testing and evaluation program for remaining VRU protection still needs to be developed.To explore the head injury mechanism from a more realistic and comprehensive perspective, 50 real-world car-to-VRU impact accident cases were selected and reconstructed numerically, through which the boundary conditions of the impact between VRUs’ head and vehicle as well as the head injury parameters during the impact events were obtained. Using linear regression model, the correlations between impact kinematics condition parameters of the head and predictions based on various head injury criteria were comprehensively investigated.The results show that a significant correlation betweenthe linear acceleration, rotational acceleration, velocity and rotational velocity and the head injury was observed. The headimpact rotational kinematics has a higher correlation with the predicted brain injury parameters than head linear kinematics, while the impact duration has no significant correlation. The result also indicated that the selection of head injury criterion needs to depend on the severity of impact boundary conditions. The current study could be an important reference for expanding the pedestrian safety tesing scenarios and determing the head injury criteria used for the head injury risk evluation.