Abstract: Target tracking is one of the most important functions in passive sonar systems. The tracked data are provided for the post-processing of target motion analysis and target recognition. Sonar operators can easily find and track the target from the bearing-time recording, by the use of the advanced cognitive model of human brain, the prior knowledge and visual trace-to-trace correlation. However, automatic multi-target tracking is very difficult to achieve in practical sonar systems, due to the complex multi-target conditions such as low signal-to-noise ratio, track cross and course varying. An automatic multi-target tracking method was proposed based on the logical association method. Broadband beam was used for data output from passive sonar. Automatic multi-target tracking could be well implemented by the use of the track association with point tracks, track evaluation management and track cross-processing. Experimental results demonstrated the effectiveness of the proposed method.