Using multi-population genetic algorithm to solve the problem of multi-path coverage is an important research direction in the field of automatic generation of test data. In order to improve the efficiency of multi-path coverage test data automatic generation, a multi-path coverage strategy combining key point probability and path similarity is proposed. Firstly, the theoretical path is divided into easily-covered, difficultly-covered, and unreachable paths. Then, the key point probability is counted through the easily-covered paths, the contribution of the individual to the generated test data is calculated by using this probability, and the contribution isusedto improve the fitness function, at the same time, the target path is sorted according to the key point probability. Finally, the test data covering the target path is generated by using multi-population genetic algorithm. After the sub-population covers the current target path during the evolution process, it continues to try to cover similar paths of the target path. The experimental results show that the proposed method can improve the efficiency of multi-path coverage to generate test data.