In the space target early warning mission, the ground-based radar needs to detect suspected missile targets from a large number of space targets, quickly and accurately judge the sudden suspicious space targets, so as to deal with the threat from space. If the target orbit parameters are analyzed and compared manually, it will not only take a long time, but also prone to misjudgment. This paper proposes to use the least square method to quickly process the real-time data of radar captured targets, extract the orbital features of different types of space targets, complete the classification of space targets, and realize the rapid early warning of missile targets. This method is not only fast, but also has a relatively low false alarm rate, so it can be used in the early warning task of early warning radar to space targets.