The restraining of water backscattering and increasing target resolution are the key technologies in underwater lidar detection. An adaptive filtering method of target echo based on recursive least- squares (RLS) algorithm is proposed. Through the feature analysis of underwater lidar detected signal, the forgetting factor in traditional RLS algorithm is improved to distinguish the target echo and water backscattering signal. The improved forgetting factor can also adapt the echo amplitude changes in different target distances. To estimate the effect of the method, the underwater detection experiments are carried out in water basin under different target distances. The results show that the backscattering signals are suppressed and the target echo are extracted with high tracking and convergence speed by using the proposed variable forgetting factor RLS algorithm. Compared with the traditional methods, the resolution of the processed target echo signal is increased and the proposed method has great advantage in the weak target echo extraction. Finally, the impact of filter order on algorithm processing results is discussed.