滑窗加权泽尼克矩特征的雷达目标识别技术

合成孔径雷达(SAR)图像的方位敏感性和相干噪声,影响SAR图像目标识别效果,针对此问题,提出了一种新的滑窗加权矩特征的雷达目标识别方法。利用三角剖分与生长切割算法得到将目标和阴影从相干噪声中分割出来的图像。根据泽尼克(Zernike)变换,计算Zernike矩,并提取滑窗加权Zernike矩作为特征不变量。最后,利用最近邻准则进行分类识别。仿真结果表明:利用滑窗加权Zernike矩作为特征向量,克服了SAR图像对方位的敏感性,有效地提高识别率,对SAR图像识别是有效的和稳健的。

Automatic target recognition of MSTAR SAR images based on sliding window weighted Zernike features

SAR images aspect sensitivity and speckle noise have a significant influence on SAR image target recognition effect. On this basis, a novel automatic target recognition method based on moving and stationary targets acquisition and recognition synthetic aperture radar(MSTAR SAR) images is proposed. At first, the target and the shadow were obtained from speckle noise by using delaunay tringulation and Growcut algorithm. Then, Zernike moments were calculated via Zernike transform, and sliding window weighted Zernike(SWWZ) moments as the feature invariants were extracted. Finally, the recognition results were obtained through utilizing the nearest neighbor criterion. The simulation results show that SWWZ moments as the feature invariants overcome the SAR aspect sensitivity and improve the recognition rate effectively, and also the proposed approach to MSTAR SAR images recognition is effective and robust.