Abstract: The problem of multiple target tracking based on particle filter in passive sonar is studied, which becomes more difficult when there are nearby targets and maneuvering targets in the observation space. First, a likelihood function considering the influence of neighboring targets is proposed. Second, the auxiliary variable is utilized to consider the measurement information in the particle sampling process. This strategy has great advantages when the actual motion of the target is mismatched with the dynamic model of the algorithm. Then, a track-before-detect algorithm is proposed, which divides the neighboring targets into a group and calculates the likelihood of the target by using the predicted states of the neighboring targets with high computational efficiency. The performance of the proposed algorithm is verified by simulation data and data collected at sea. Compared with other multi-target particle filter track-before-detect algorithms, the proposed algorithm has better tracking performance, especially in the presence of neighboring and maneuvering targets.