The work aims to solve the optimization problem of three-dimensional container loading space utilization in the cargo hold of unmanned aerial vehicles (UAVs), break through the space utilization bottleneck of the traditional empirical container loading method, and achieve efficient loading under multiple constraints. In combination with the requirement of high support degree for loading goods by unmanned aerial vehicles, a highly stable mathematical model was constructed. A hybrid optimization method combining genetic algorithm and simulated annealing algorithm was adopted. The initial population was generated through chaotic mapping. Then, two-stage coding was used, combined with the dynamic spatial segmentation method for loading optimization. Finally, the optimal loading scheme was iterated out. Simulation experiments were conducted on 390 parts of 20 types in an automotive parts enterprise using the proposed hybrid algorithm. Compared with the original algorithm under the premise of meeting the support constraints and other constraints, the space utilization rate was increased by 5%, indicating that the packing effect was significantly better than other algorithms. In conclusion the improved algorithm has high loading efficiency and strong stability, providing an effective optimization method for the cargo hold loading problem of unmanned aerial vehicles. It can ensure an optimal loading effect for all kinds of goods, solve the space utilization problem existing in the empirical packing method of a certain enterprise, and at the same time provide a reference for the future research on the packing problem of unmanned aerial vehicles, and has a good application prospect.