Under the early warning of two passive satellite-borne sensors, the exponential weighted recursive least square method was applied to the burnout states estimation of ballistic targets. Due to the weighting factor can characterize the local quasi-linearity of the boost phase trajectory of the target, so to some extent the dilemma that generic polynomial cannot handle the whole boost phase of the target can be overcome. Through a dynamic analysis, the kinematic characteristics of boost phase target along the orientation of vertical surface were explored. On that basis, a novel boost phase motion model is proposed, which is more accurate than traditional planar motion model. Simulation results show that the proposed approach is superior to the traditional ones.