To reduce the complexity of target search and the randomness of environmental features in unknown environment, this paper proposes a subareabased perceptual adaptive target search algorithm to solve the traditional comb search pattern or the problem that in global optimization track, autonomous underwater vehicles (AUVs) cannot flexibly adapt to the environment. The main feature of this method is to design the optimal trajectory planning in realtime according to the characteristics of the environment target in the AUV vision and to use the Bayesian estimation to complete the target localization. When there is no target, use the subregional grid values and lock the task regional trajectory planning, and the two modes alternate to improve the search flexibility. Simulation results show that compared with the traditional search mode or the global optimization track, this method enhances the environment adaptability and improves the target search efficiency.