In radar automatic target recognition(RATR),the high-resolution range profile(HRRP)has garnered considerable attention owing to its minimal computational demands.However,radar HRRP target recognition still faces numer...In radar automatic target recognition(RATR),the high-resolution range profile(HRRP)has garnered considerable attention owing to its minimal computational demands.However,radar HRRP target recognition still faces numerous challenges,primarily due to substantial variations in the amplitude and distribution of HRRP scattering points because of slight azimuthal changes.To alleviate the effect of aspect sensitivity,a novel multi-frame attention network(MFA-Net)comprising a range deformable convolution module(RDCM),multi-frame attention module(MFAM),and global-local Transformer module(GLTM)is proposed.The RDCM is designed to adaptively learn the distance of scattering center migration.Subsequently,the MFAM extracts consistent features across different frames to alleviate the influence of power fluctuation.Finally,the GLTM allocates attention between global and local fea-tures.The feasibility and effectiveness of the proposed method are validated through simulation and experimental datasets,and the recognition rate is enhanced by more than 3%compared to the state-of-the-art methods.展开更多
基金The National Natural Science Foundation of China(No.62388102)the Natural Science Foundation of Shandong Province(No.ZR2021MF134).
文摘In radar automatic target recognition(RATR),the high-resolution range profile(HRRP)has garnered considerable attention owing to its minimal computational demands.However,radar HRRP target recognition still faces numerous challenges,primarily due to substantial variations in the amplitude and distribution of HRRP scattering points because of slight azimuthal changes.To alleviate the effect of aspect sensitivity,a novel multi-frame attention network(MFA-Net)comprising a range deformable convolution module(RDCM),multi-frame attention module(MFAM),and global-local Transformer module(GLTM)is proposed.The RDCM is designed to adaptively learn the distance of scattering center migration.Subsequently,the MFAM extracts consistent features across different frames to alleviate the influence of power fluctuation.Finally,the GLTM allocates attention between global and local fea-tures.The feasibility and effectiveness of the proposed method are validated through simulation and experimental datasets,and the recognition rate is enhanced by more than 3%compared to the state-of-the-art methods.
