The large field optical telescope is an important equipment for the search of medium-high orbit targets. However， stellar targets also appear in the search images. Thus， the identification and suppression of stellar targets is a necessary step in the detection of medium-high orbit targets. Considering the influence of the sky area near the galactic plane， difference in exposure time， and cloudy occlusion， the variation in the star field density within the image is very large. The traditional star identification method has limitations in the calculation accuracy and real-time performance， which leads to the occurrence of false alarm of stars and calculation timeout. To solve these problems， this study proposes a method of star identification and suppression based on the relative invariance of inertia coordinates in time domain. First， the mathematical transformation relationship between the horizon coordinate system and the inertial coordinate system is derived， and the star identification model is constructed accordingly. Then， the time domain relative invariance of the inertial coordinates of the stellar target is quantitatively analyzed under different static system errors. Finally， the star identification and suppression algorithm was verified via simulation and experiment. As a result， the maximum relative difference in the inertial coordinates of stars is 0.51" （right longitude） and 0.16" （declination） when the time interval is 10" and the static system error is 10". Thus， the time domain relative invariance meets the requirements regarding star identification， and the corresponding process is completely independent of star field density. The proposed method was also verified considering 100 measurement images of medium-altitude orbit objects， and no false alarm of stars and missing detection phenomenon of medium-altitude orbit objects were observed.