This paper presents a methodology for geometric personalization of numerical human body models, specifically focusing on adapting the 50th percentile male THUMS model to represent female anthropometry for pedestrian-vehicle collision analysis. Road accidents involving pedestrians represent a significant social and economic problem worldwide, with pedestrians being one of the most vulnerable road users. The study developed a morphing method using the Dual Kriging algorithm to transform the external and internal geometry of the base model while maintaining numerical stability. The process involved using a 3D female surface model as a reference, mapping control points, and ensuring anatomical accuracy especially in the thoracic region. The personalized models were validated through simulated collision tests with a passenger vehicle at 40 km/h, which demonstrated that the models maintained expected kinematic behavior while reflecting the anthropometric differences. Results confirmed that geostatistical algorithms can effectively personalize numerical human models, providing a foundation for future injury analysis considering population diversity.