基于小波域中心矩特征的SAR图像识别

为克服合成孔径雷达(SAR: Synthetic Aperture Radar)图像的方位敏感性和平移敏感性给识别带来的困难, 提出一种基于二维离散小波变换与中心矩特征提取的SAR图像目标识别方法。该方法通过对图像的二维离散小波分解提取低频子带图像, 同时提取具有平移不变性的中心矩作为特征向量, 利用支持向量机进行目标分类和识别。实验结果表明, 该方法在有效抑制噪声的情况下, 很好地克服了SAR图像对目标方位的敏感性, 在减少计算量的同时具有较高的识别率。

SAR Images Target Recognition Based on Wavelet Domain Central Moments Feature

In order to overcome difficulty bright by the azimuth sensitivity and translation of SAR (Synthetic Aperture Radar) image,we presents a method of synthetic aperture radar image recognition based on two-dimension discretewavelet transform and central moments feature extraction. After two dimension wavelet decomposition of the image, feature extraction is implemented by pickingup the low-frequency sub-band image.And the central moments with translation-invariant property is extracted as feature vector.Support vector machine is used to classify the feature vector for target recognition. Experimentresults show that the method is an effective method that can reduce calculation and enhance the recognition.