Abstract: It is not easy to establish statistical test model in correlated non-Gaussian clutter backgrounds when parameter is stochastic or unknown. Based on the particle filtering, the proposed method calculates the likelihood function by transforming the integral operation to the sum operation according to the probability distribution function of unknown parameter, and a general radar target likelihood-ratio detection model is established. This method solves the difficulties that it is unable to establish statistical test model in correlated non-Gaussian clutter backgrounds. Taking SG-Alpha stable distribution as the model of correlated non-Gaussian clutter, a statistical test method based on particle filtering is established. Moreover, simulations show that the detection performances of the proposed method are better than the traditional method both in correlated Gaussian and non-Gaussian clutter backgrounds.