The work aims to research a multi-container loading optimization algorithm, so as to solve the problem that most of the current packaging algorithms do not consider the loading order constraints and cannot effectively solve the problem of practical application of air cargo loading. First, the K-Means algorithm was used for cargo pre-allocation, and the cargoes with the same clustering characteristics were assigned to the same container. Then, the extreme point order was obtained by extreme points (Eps), and the genetic algorithm was used to optimize the layout scheme of each container. 160 cargo data of an airport logistics company were used for experiments and compared with the continuity strategy, which proved that the Eps loading method with pre-allocation strategy could effectively avoid the low utilization of individual container and make the population variance of container utilization drop to 0.51. Considering the constraints of cargo loading order, the algorithm can realize the rational allocation of cargo in multi-container loading optimization and has better engineering applicability.