Vehicle–pedestrian impact is one of the major traffic accident types all over the world. In these accidents, head injury is the main factor that causes the pedestrian death. However, numerical human models that have been developed cannot be applied effectively in these accidents either because the head structure is not precise enough to show the local damage or because the model is too complex to ensure the computation efficiency. For instance, the Head Injury Criterion (HIC) of the TNO Hybrid III dummy model could be used for the prediction of skull fracture, and the contact point between the head and the vehicle could give an indication of where the injury occurred. However, the dummy model is not able to show the local stress and damage of the head that is shown in accident autopsy reports. Therefore, the anthropometry-based finite-element head and multibody knockdown pedestrian models have been developed. The knockdown human model cannot only simulate precisely the head damage and more accurate damage position but also improve the computing efficiency. Additionally, the parameter customisation technology is applied for the pedestrian model in order to get rid of using the standard pedestrian model in all cases. The purpose of this paper is to prove the validity for the new pedestrian model applied in accident simulation and optimisation. By reconstructing real-world vehicle–pedestrian impact accidents, the performance of the modelling method is evaluated. The results show that the method is very effective in finding an optimum solution and receiving enough exact head injury value.