The invention discloses a luminosity-curve-based real-time online characteristic recognition method for a spatial high-orbit small-sized target so that problems of spatial high-orbit target characteristic recognition and motion estimation can be solved and a defect that the existing method can not be converged under luminosity observation can be overcome. On the basis of analyses of the dynamic characteristics and luminosity observation data characteristics of a high-orbit target, a novel idea for characteristic recognition is realized by combining the attitude estimation of the target. The method is a parallel fusion algorithm established based on a target attitude kinematics model, so that the convergence and convergence speed of the algorithm are ensured; and characteristics of the target shape, the dimension, and the material parameter and the like can be identified effectively. The method is established based on a multi-model hybrid estimation framework; and the target attitude estimation and characteristic recognition are realized synchronously in an optimized manner.