基于多小波子带加权判别熵的SAR目标ICA特征提取及识别

SAR target feature extraction and recognition based on multi-wavelet sub-band weighted discrimination entropy ICA

传统小波独立分量分析(ICA)提取合成孔径雷达(SAR)目标特征时大都采用单一的小波基函数,并且仅利用小波分解低频子带数据进行ICA处理,而忽略了高频子带信息。针对这一问题,采用多类小波基函数对SAR目标图像进行分解;针对得到的所有低频和高频子带数据,引入子带加权的判别熵准则,结合现有的小波ICA算法,提出多小波子带加权判别熵的SAR目标图像ICA特征提取算法。采用MSTAR实测SAR目标图像数据,根据提出算法进行特征抽取,利用最近邻准则进行SAR目标识别。识别结果表明提出算法优于仅利用小波分解低频子带ICA算法。

Generally, single wavelet basis function and low frequency sub-band of the signal are used to perform Independent Component Analysis （ICA） in wavelet domain for Synthetic Aperture Radar （SAR） target feature extraction while high frequency sub-bands are ignored. For this defect, SAR images were decomposed utilizing multi-wavelet basis function. Then a new feature extraction method was proposed by multi-wavelet sub-band ICA according to sub-band weighted discrimination entropy criterion on the basis of the general wavelet-ICA algorithm. The SAR target recognition experiment was performed on the nearest criterion using features extracted by the new algorithm with MSTAR dataset. The experimental results show that the proposed algorithm is superior to the traditional wavelet-ICA algorithm.