Singularity disposing, salient point gaining and random error weakening are key processes for estimating precision during the data processing for ballistic trajectory. Firstly, the predominance of wavelet transform on dealing with singularity, salient point and random error, was used to conduct a decomposition, fusion and reconstitution of the observation data base on wavelet transform , which can eliminate singularity, identify salient point and weaken random error. Secondly, the B spine function with knots free distributing was used to describe the missile movement state, which makes the principle of the trajectory determination process become a nonlinear optimization problem of multi model fusion with the parameters of missile trajectory expression coefficients and observation system error coefficients. By introducing the model structure to determinate the selection rule of optimal fusion weight, the optimal estimation for the to be estimated parameters was obtained, and more the precise trajectory determination was gained. Finally, the simulation calculation shows that this method has some advantages in disposing singularity, gaining salient point and weakening random error, which can reduce the calculation amount and increase the ultimate precision of trajectory determination effectively.