The Rao test of weak signal in non-Gaussian clutter is addressed. Both the models of Gaussian mixture distribution and Alpha-stable distribution are used to simulate the clutter distribution of airborne early warning radar. The improved expectation maximization algorithm [4] and iterative algorithm are used to estimate the parameters of Gaussian mixture distribution and Alpha-stable distribution respectively, and then the Rao decision rule under non-Gaussian clutter environment is presented. Theory analyses and simulation results indicate that the detection performances of Rao test based on non-Gaussian clutter hypothesis outperform that of Rao test based on Gaussian clutter hypothesis.