The high-resolution range profile (HRRP) presents significant potential for few-shot aerial target recognition due to its low complexity and rich structural information compared to conventional techniques. Nevertheless, few-shot HRRP recognition encounters two primary challenges: intraclass variations resulting from HRRP's sensitivity to prior information (such as azimuth, elevation, and distance) and overfitting in few-shot scenarios. This article proposes a prior information-assisted few-shot HRRP classifier (PFHC) to address these challenges. The PFHC integrates a prior information-assisted gated recurrent unit (PA-GRU) and task-wise shrinkage quadratic discriminant analysis (TS-QDA). The PA-GRU is the feature extractor, designed to learn the latent relationships between prior information and intraclass variations. Following feature extraction, TS-QDA serves as the few-shot classifier, leveraging feature covariance to mitigate overfitting while maintaining stability through task-wise shrinkage. Experimental results indicate that the PFHC effectively captures discriminative features for the HRRP recognition using only 10% of the parameters required by the conventional feature extractors. Moreover, the PFHC demonstrates superior recognition accuracy across most few-shot task settings and scenarios compared to several state-of-the-art inductive few-shot classifiers. For instance, the PFHC achieves an accuracy improvement of at least 4.24% and 8.05% in the “5-way 20-shot” scenario and generalized scenario with 3-way training and 20-shot, respectively.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Prior Information-Assisted Few-Shot HRRP Recognition Based on Task-Wise Shrinkage Quadratic Discriminant Analysis


    Beteiligte:
    Li, Jixi (Autor:in) / Guo, Weiwei (Autor:in) / Wei, Feiming (Autor:in) / Zhang, Tao (Autor:in) / Yu, Wenxian (Autor:in)


    Erscheinungsdatum :

    01.12.2024


    Format / Umfang :

    4915057 byte




    Medientyp :

    Aufsatz (Zeitschrift)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch





    Target Recognition of Radar HRRP Using the Envelope Reconstruction

    Zhang, Pengfei / Chan, Li / Zhou, Hongxi et al. | Springer Verlag | 2017


    Target Recognition of Radar HRRP Using the Envelope Reconstruction

    Zhang, Pengfei / Chan, Li / Zhou, Hongxi et al. | British Library Conference Proceedings | 2018


    Radar HRRP Target Recognition Based on Hybrid Quantum Neural Networks

    Liu, Xin / Zhou, Daiying / Huang, Qiuyan | IEEE | 2025