In view of the influence of strong heterogeneous goods on loading efficiency and the influence of loading results on vehicle stability in logistics loading, a three-dimensional packing algorithm with the goal of maximizing the utilization rate of compartment space and the minimum offset of vehicle center of gravity was proposed based on anthropomorphic loading strategy and improved genetic algorithm. The algorithm considered a variety of realistic constraints such as cargo type, volume, load-bearing capacity, non-overlapping, etc., and integrated the KD-tree structure to optimize the running speed of the algorithm, and useed Bayesian parameter optimization to obtain the best parameters of the algorithm. The algorithm was applied to a logistics enterprise to optimize the loading of a certain batch, and the maximum space utilization rate was 93.63%. The Bischoff & Ratcliff standard example was used for testing, and the space utilization rate was up to 93.03%. It has been verified that the algorithm can effectively solve the packing problem of strong heterogeneous goods and provide a better loading solution for logistics enterprises.