Explainable Hybrid Decision Level Fusion for Heterogenous EO and Passive RF Fusion via xLFER

This paper presents an explainable late-stage decision fusion model for Electro-Optical (EO) and Passive Radio Frequency (P-RF) target detection via hybrid Explainable AI model. Explainable insights that are intuitive and empirical are provided by counterfactual explanations at the early stage of data flow, with a traditional algorithm, decision tree (DT), handling late-stage fusion. Results show that at both the local and global level, the DT explainability of fusion methods provides insights for EO and P-RF fusion methods at each level of fusion The usage of Histograms, Wigner-Ville Distribution (WVD) and Continuous Wavelet Transform (CWT) for the novel use of P-RF data provided insights into the eXplainable Late-stage Fusion of Electro-optical and Radio-Frequency (xLFER) usage of the modality for target detection. While WVD and CWT have been used extensively in RF signal processing, their use in P-RF data for target detection feature extraction has not been documented to our knowledge, nor with a hybrid Explainable AI model.