LIME-Assisted Automatic Target Recognition with SAR Images: Towards Incremental Learning and Explainability

Integrating an automatic target recognition (ATR) system into real-world applications presents a challenge as it may frequently encounter new samples from unseen classes. To overcome this challenge, is necessary to adopt incremental learning, which enables the continuous acquisition of knowledge while retaining previous knowledge. This paper introduces novel, multi-purpose interpretability metric for ATR systems that employ synthetic aperture radar (SAR) images. The leverages Local Interpretable Model-Agnostic Explanation (LIME) algorithm, enhancing human decision-making by providing secondary measure alongside conventional classification score. Additionally, proposed employed analyze robustness Convolutional Neural Networks (CNNs) examining impact features and irrelevant background correlations on results. Finally, we demonstrate effectiveness in context learning. By utilizing metric, select exemplars learning scenario, resulting improved performance showcasing application potential our methodology. network fine-tuned sequentially with unknown recognized Openmax classifier old known classes, are selected based metric. approach demonstrated using publicly available MSTAR dataset.