Cuckoo search algorithm (CS) is a valid bio-heuristic algorithm, which has been extensively applied to solve the optimal problem in actual engineering projects, due to the advantages of simplicity, few parameters and easy implementation. In order to improve the population diversity and global search efficiency of the standard CS algorithm, the different mutation processes of an improved difference evolution algorithm was introduced into the cuckoo algorithm. In the different mutation processes, the multi-strategy associated with random walks method of the CS algorithm was used to optimize the host discovery process. With the multi-strategy difference mutation operation, the diversity of the cuckoo population was improved in the process of the cuckoo searching. Meanwhile, in the improved cuckoo searching, the queue optimization mechanism was added to the new solution selection, combining with the greedy algorithm to reduce the attraction problem of the undesirable solution and speed up the search process. In addition, the improved cuckoo algorithm with multi-strategy different mutation processes was applied to particle filtering. The particles were characterized with the cuckoo nests, by simulating the process that the cuckoo groups searched the nests to optimize the particle distribution. The experiment result showed that the improved particle filter can improve the prediction accuracy of particle diversity and nonlinear system state, and it can keep a good robustness and stability in the case of the particle number decrease.