In human vulnerability assessment, the wound characteristics are derived from a single soft media penetration experiment, and there is a lack of mathematical model or algorithm to accurately describe the bullet motion (consider bullet rolling) and temporary cavity variation during bullet penetration into different soft media. This paper derives a bullet motion and cavity expansion-contraction model for bullet penetration into soft tissue; Established a dynamic wound reconstruction model based on neural networks that considers tissue differences. Assessment of damage to tissues using the Abbreviated Injury Scale; Developed software for assessing human vulnerability based on dynamic wound reconstruction. Research results show that the bullet motion model and the cavity expansion-contraction model can predict the characteristic volume of the wound and the temporary cavity profile changes during bullet penetration more accurately; the maximum temporary cavity diameters of the muscle wound, the cardiac wound, and the muscle-cardiac-muscle coupling wound are 183.6, 158.06, and 174.74 mm respectively, and using the cavity in a single target as the basis for human injury assessment will introduce errors. The process of bullet penetration into soft tissue can be accurately described based on a predictive model that considers tissue differences. This paper provides the model that improves the accuracy of human injury assessment compared to existing penetration models.