A spline filtering algorithm based on interacting multiple model algorithm (IMM-SF) was proposed to estimate Line-of-sight (LOS) rate in strap-down seeker system. Based on the mapping relation between body LOS and inertial LOS, the inertial LOS angles were obtained and used as virtual observed data of filters. A set of process noise models were set up, and every model worked independently based on the spline filter. The valuation output of IMM-SF was the weighted composition of the valuation outputs of all parallel spline filters. By using this method, the unknown maneuver models of the target were no longer indispensable, which makes the method more convenient to use. Process noise could be adaptively adjusted under the configuration of IMM. The Monte-Carlo simulation results show that the LOS rate can be estimated effectively based on SF-IMM, which can also improve the accuracy of estimation.