Generalized Multivariate Decision Theory: A Novel Target Detection in Clutter Using Joint Bayesian & Generalized Likelihood Tests in Neyman-Pearson De

A novel Decision Theoretics algorithm based on cascaded generalized and Bayesian likelihood tests in a Neyman-Pearson detection was demonstrated under Phase-I. OKSI demonstrated EO/IR subpixel target detection, identification, and countermeasure mitigation. The algorithm was tested with the airborne HYDICE sensor and the Forest Radiance data set. Detector ROC curves were developed to establish the detection probability (Pd) under a constant false alarm rate (CFAR) for various sensor signal-to-noise (SNR) values. Background statistics were established using a novel in-scene technique by first creating the target subspace, and second, projecting the scene into an independent subspace. Specific target subspaces can be generated from the Red Team-provided data, and engagement scenario simulations. During the proposed Phase-II, the operability of the algorithm will be tested under more stressing and BMDS-relevant engagement scenarios, with real sensor operating factors such as PSF, reduced number of bands, broader bandpass, etc. OKSI will work with Project Hercules Blue and Red Teams and the Corporate Clutter Working Group (CCWG) to establish realistic threat signatures and demonstrate target detection, discrimination, and countermeasure and clutter mitigation capabilities. Also, working with the Decision Architecture (DA) Team, OKSI will contribute to the physics-based discrimination and EO/IR sensor integration. The proposed technology is applicable to all phases of the BMDS and the various support sensors such as DSP, STSS, SBIR High, CB, HALO-I/II, and AIRS, as well as the KV onboard-sensors for plume-to-hardbody handover and aim point selection (boost, exo-atmpshperic, and terminal intercept), and finally, in closely spaced object conditions.