Automatic Target Recognition for SAR Images Based on the Combination of GCN and CNN
The automatic target recognition(ATR)technology based on convolutional neural network(CNN)for synthetic aperture radar(SAR)has attracted much attention in recent years,becoming a research hotspot in the field of SAR image interpretation.However,these methods primarily utilize the amplitude information of SAR images and only extract features from local regions.Given that targets in SAR images are typically regarded as the coherent superposition of scattering centers,these targets exhibit complex structures and rich contextual information.It is difficult to fully cap-ture the global information around the target by relying only on CNN,which may affect the recognition accuracy.There-fore,to further improve the recognition performance,this study introduces the graph convolutional network(GCN)and proposes a SAR ATR method combining GCN and CNN.This method first utilizes traditional CNN to extract local fea-tures related to the amplitude of SAR images,and then employs GCN to extract global features by constructing graph data.Additionally,a multi-scale GCN is designed to enhance the interpretation ability of the graph data by fusing fea-tures from different scales.During the model training phase,the label smoothing technique is employed to alleviate the overfitting.Through an end-to-end training strategy,the joint optimization of GCN and CNN parameters ensures a high-precision SAR image target recognition.Finally,experimental results on the MSTAR and OpenSARship datasets demon-strate that the proposed method outperforms the existing techniques in terms of recognition performance and exhibits superior generalization capability.
万方数据
维普
