Fast passive infrared detection and trajectory estimation of UAVs using a physics-based GLRT and hierarchical superparameter mapping

Passive infrared (IR) sensing is attractive for surveillance and target tracking because it does not emit detectable active interrogators, can operate in day/night conditions, and provides thermal contrast that complements active modalities such as radar. However, detecting dim moving thermal targets in low signal-to-noise ratio (SNR) regimes remains difficult. Conventional IR workflows frequently rely on image-based tracking after detection, while direct low-SNR detection and parameter estimation through a joint maximum-likelihood or generalized likelihood ratio test (GLRT) can be computationally expensive and sensitive to local trapping in noisy rugged optimization landscapes. In this paper we formulate a physics-based GLRT for passive IR sensing of an unmanned aerial vehicle (UAV) using a forward radiometric model that relates detector thermal power measurements to the UAV thermal power, initial position, and velocity. We then introduce a hierarchical superparameter mapping (HSM) strategy that replaces a seven-parameter joint optimization by a sequence of smaller estimation problems. The method first estimates detector-level superparameters through ordinary least squares, then reconstructs the physical parameters through low-dimensional inverse steps with limited feedback between branches. In simulation, the HSM pipeline retains the accuracy of direct maximum-likelihood estimation (MLE) in successful cases while substantially improving latency and robustness in noisy conditions. For the reported test set, average runtime decreased from 9.4s for direct MLE to 113ms for this implementation of HSM.