In order to overcome the inadequacy of conventional RCS statistical models describing stealth target fluctuation, a new RCS statistical modeling method based on Gaussian mixture density model is presented. The two-order GMDM and distribution model are established respectively based on the simulation data of typical stealth target. The results show that the forward, side and backward fitting errors of GMDM are respectively 4.74%, 12.34% and 1.01%, while those of the model are 44.5%, 18.65% and 13.21%. And simultaneously when fitting order in number exceeds four-order, the fitting error of GMDM remains below 5% steadily, which can satisfy the precision demand of radar target simulation.