Robust Topological Association Method Based on Graph Theory Matching

Robust target association is a key task in distributed multi-radar fusion, especially when systematic errors, false alarms, and missed detections coexist. To improve association robustness under such conditions, this paper proposes a graph-theory-based topological matching method. First, a relative-position graph is constructed for each target, and an ordered topological relationship matrix with rotation-invariant properties is generated. Second, a truncated-sector neighborhood constraint is introduced to restrict candidate matches and reduce redundant calculations. Third, rotation-translation compensation is combined with a positiondifference- based topological metric, so that the influence of radar systematic bias can be reduced before association testing. Finally, the association decision is formulated as a hypothesis-testing problem, and the corresponding threshold selection rule is given. Monte Carlo simulations with an 85% detection probability, a 10 -3 false-alarm probability, and different error/spacing settings show that the proposed method maintains an association accuracy of about 95% in large systematic-error and falsealarm/ missed-detection scenarios. For dense formations, a global matching refinement is further introduced to overcome the degradation caused by small inter-target spacing. The results indicate that the method provides a useful balance between robustness, computational efficiency, and engineering applicability.