一种SAR图像车辆目标鉴别特征及其提取方法

该文引入差分盒维法来计算高分辨率SAR图像车辆目标的一种新的鉴别特征,即间隙度特征,可以用来定量评估车俩目标感兴趣区域内像素幅度的不规则程度和间隙尺寸,以此消除杂波虚警.基丁散射中心理论分析了车辆目标和自然地物后向散射强度分布的差异性,并从理论上推导出间隙度特征具有对相干斑噪声不敏感的特点,由此构成了SAR图像车辆目标鉴别处理的一个尺度不变特征.采用MSTAR车辆目标数据和背景杂波数据检验了所提特征的鉴别性能,并与Hausdorff维数的鉴别性能做了比较,结果显示间隙度特征具有较好的鉴别性能,可以去除大部分的自然地物虚警和非车辆类人造目标干扰,鉴别虚警率较低.

New feature for vehicle target discrimination in SAR imagery

The differential box-counting algorithm is introduced to calculate a new discriminating feature named Lacunarity, which is used to distinguish vehicle targetfrom naturalclutter in high-resolution SAR imagery in thispaper. Lacunarity feature can be used to estimate quantitatively the variation, irregularity and gap size of pixel s intensity of candidate targets. Based on the theory of scattering center, it can be shown that the vehicle image presents more irregularity and larger gaps than natural terrain s image. Moreover, lacunarity is robust to speckle noise and is stable under changes in intensity. Finally, the realvehicle targetdata and natural terrain sdata inMSTAR database are applied to test the above algorithm. The discrimination performance using lacunarity is compared withHausdorff dimension. The result shows that lacunarity is a good discriminating feature, which can eliminatemost false alarms from natural terrains andmost interference from theman-made targetswith low false alarm probability.