In view of characteristics of the scale, multi product and multi machine features of a TFT-LCD cell assembly stage, the learning and deterioration effects were introduced. The minimization of the maximum completion time, the total wait time and the weighted delay time of the machine were taken as objective functions, and a multi-objective TFT-LCD cell scheduling model was built. Based on the two segments and IMM encoding, the multi-objective cuckoo algorithm combined with the Pareto set with dual championship, the dynamic elimination rules and the cluster distance density evaluation index to solve the multi-objective TFT-LCD cell scheduling problem. The simulation results show that the cuckoo algorithm is superior to the elitist reservation greedy decoding genetic algorithm, the process expectation shortest completion time scheduling regulation, etc. Through experiments, the effects of different learning rates and deterioration factors on scheduling results were analyzed.