The present study describes the development of a refined finite element model of the human pelvis. The objectives of this research work were to: Statistically study the human pelvis geometry, and develop a parameterized model. Mechanically validate the model with regard to the available in-house experimental data. Model the injury mechanisms observed in the experimental studies. The significant dimensions of the pelvis were identified by statistical analysis of the pelvis geometry based on the Reynolds et al. data . Those dimensions were used to classify the in-house tested pelves. An interpolation technique (Kriging [2, 3, 4, 5, 6, 7 and 8]) was then used in order to distort a reference mesh and adapt its geometry to the measured geometry of the tested pelvis. The mechanical validation of the model was carried out by comparing numerical and experimental results, and the influence of the geometrical variations on the behavior of the pelvis was thus assessed. Some fracture phenomena were then modeled and the model was validated using injuries observed in experiments in terms of displacements and rupture mechanisms (beginning and propagation of the fracture).