二维变分模态分解在SAR图像特征提取及目标识别中的应用

Application of Bidimensional Variational Mode Decomposition in SAR Image Feature Extraction and Target Recognition

将二维变分模态分解(BVMD)应用于合成孔径雷达(SAR)图像特征提取及目标识别,经过BVMD分解后,产生原始SAR图像的若干分解成分,可更有效地描述目标的全局和局部细节信息。首先,决策过程利用支持向量机(SVM)对原始图像及分解成分进行独立分类;然后,基于线性加权融合对它们的结果进行处理,根据最终的决策变量获得目标类别决策;最后,采用MSTAR数据集开展测试,分别在标准操作条件、俯仰角差异和噪声干扰条件下进行实验,并与现有几类方法进行比较,结果验证了方法的有效性。

The Bidimensional Variational Mode Decomposition(BVMD)is applied to Synthetic Aperture Radar(SAR)image feature extraction and target recognition.Multiple components are generated after BVMD of the original SAR image which can effectively describe the global and local information of the target.In the phase of the decision-making the Support Vector Machine(SVM)is adopted to classify the original SAR image and its decompositions separately.Afterwards linear weighted fusion is employed to combine their results.Finally based on the fused results the target label of the test sample can be determined.The proposed method is tested based on the MSTAR dataset respectively under standard operating condition with pitch angle variation and with noise interference which are compared with several existing SAR target recognition methods.The experimental results verify the effectiveness of the proposed method.