An improved Interactive Multiple Model Strong Tracking Cubature Kalman Filter (IMMISTCKF) algorithm is proposed to address the tracking problem during maneuvering flight of non-cooperative Boost Glide Vehicles (BGVs). A system motion model is established based on the flight characteristics of BGVs, and an observation model is developed according to the measurement principles of space-based optical sensors. A framework of the algorithm based on Interactive Multiple Models (IMM) is designed, which integrates maneuver models with different maneuver frequencies to achieve adaptive filtering of maneuvering processes without prior information. Fading factors are introduced in the algorithm to enhance its stability. Simulation comparisons of IMMISTCKF with CKF, STCKF, and IMMCKF algorithms demonstrate that the proposed algorithm provides higher accuracy in position and velocity estimation for hypersonic maneuvering target tracking.