Addressing the challenges of real-time precise localization and dynamic threat assessment of low, slow, small targets (LSS) in multi-domain, multi-dimensional environments, this paper proposes a method combining the unscented Kalman filter (UKF) with dynamic Bayesian networks (DBN). These challenges include tracking LSS targets’ unpredictable motion and assessing their threat levels in real time. A mathematical framework for real-time localization is established based on the motion characteristics of LSS targets. Two approaches are employed to deduce real-time state information: one based on UKF filtering estimates and the other on traditional single-point calculation. The DBN model is utilized for dynamic inference of the target’s threat level. Simulation results demonstrate that this method accurately captures temporal changes in threats, exhibits good robustness, and provides scientific support for rapid decision-making in modern air defense.