Fast Prototype Refinement for HRRP Target Recognition with Missing Aspects

Résumé fourni par la source

High-resolution range profile (HRRP) based radar automatic target recognition (RATR) has attracted significant attention due to its low signal processing complexity and the rich structural information it provides about targets. However, HRRP is highly aspect-sensitive, making it challenging for a trained classifier to perform well on new aspects not encountered during training. To address this issue, we propose a fast prototype refinement method designed to enhance the representativeness of pre-trained class prototypes using test HRRPs from unseen aspects in an unsupervised manner. Our framework comprises two iterative procedures: transfer matrix estimation and prototype update. The transfer matrix estimation extracts transferable features from the test HRRPs, while the prototype update improves the discriminability of class prototypes based on these features. To do so, we introduce a generalized mixture Gaussian model with a nonlinear distance to measure similarities between class prototypes and test HRRPs. An Expectation-Maximization (EM) algorithm is then employed to maximize the log-likelihood of the generalized mixture Gaussian model given test HRRPs. Due to the convergence of the EM algorithm, these refined prototypes are robust for HRRP-based RATR in scenarios with missing aspects.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.