The characteristics of gathering fresh agricultural products, such as multiple nodes, scattered demands, cargo restrictions, and product perishability, are considered in studying split delivery route optimization. The basic problem of the gather of fresh agricultural products are described through node layout, network demand, vehicle routing and load requirements, product characteristics, and cost factors. The total gathering cost is modeled based on three aspects: refrigeration, cargo damage, and fuel consumption. Constraints are designed around split delivery, flow balancing, and vehicle scheduling to construct an optimized gathering route model. Using a genetic algorithm, the fitness function is calibrated by serial number sorting, and roulette wheel selection is improved. An elite strategy is employed for operator analysis and genetic operations, leading to a proposed solution method. An empirical analysis of orange gathering in a township in Hunan Province demonstrates the results: the optimized total gathering cost decreases by 19.34%, with the cargo damage cost reducing the most, by 1550.86 yuan. The total number of gathering routes is reduced by 1, and the number of splitting nodes decreases from 5 to 3, reflecting a deeper split by demand. The total driving distance and time are reduced by 72.4 km and 2.06 hours, respectively, showing a significant overall optimization effect. The research provides valuable insights for improving the efficiency of fresh agricultural product collection.