As the basis of decision-making, distributed radar array-based Situation Awareness (SA) plays a critical role in UAV swarm combats. To improve the accuracy of SA, a modified Endsley three-level SA model is proposed, which focus on distributed radar signal processing and fusion. Firstly, a consistency maintenance algorithm is proposed to deal with the inconsistency of the situation information within swarm. Secondly, an identification model of the topological structure and nodes is constructed in SA cognition layer. Finally, the interactive multiple model algorithm and analytic hierarchy process are introduced to estimate the states of units and the threat level of the whole swarm. Besides, the effectiveness of the method proposed in this paper is verified by numerical simulation.