Abstract: Based on static measurement data of microwave anechoic chamber,this paper constructs the dynamic radar cross section(RCS) time series for a ballistic target in typical scenarios.The simulation indicates that the time series have the characteristics of nonstationarity,correlativity and quasi-periodicity.The effect of direct power spectrum estimation of precession period is not significant.This paper uses B-spline function,generalized autoregressive conditional heteroskedasticity(GARCH)model and autoregressive-moving average(ARMA)model to constitute a cascaded filtering model,in which the B-spline function represents the tendency term,GARCH model represents the volatility clustering term,and ARMA model represents the noise corrupted sinusoidal term.We demonstrate theoretically that the sinusoidal term is equivalent to an ARMA sequence with slowly varying parameters.The results show that the method can effectively extract the precession period of the ballistic target.