Abstract: A novel method based on Gaussian particle filter (GPF) for underwater target tracking was presented aiming at the uncertain and fuzzy information of moving objects obtained by sonar sensor in complex underwater environment, which takes account of the inherent characters of sonar sensor. A first-order autoregressive-process equation was used as the support of state transition of moving object according to the particle filter theory. A measurement model combining the object region with its moment invariants was designed. The problem of particles weight selection was solved and the resample of traditional particle filter was avoided. The correct rate of object tracking under complex background was improved. The complete procedure of underwater object tracking based on Gaussian particle filter was displayed. The results show that the advanced method has satisfactory robustness and real time property. It is a feasible and -effective-way for object tracking in complex underwater environment．

Key words: Gaussian particle filter, object tracking, sonar detection, moment invariants, particles weight