SAR场景图像中车辆目标的自动识别

提出一种在合成孔径雷达(synthetic aperture radar,SAR)场景图像中进行自动目标识别的方法,并针对识别流程中分类阶段对目标方位角敏感的问题,基于相关滤波器理论与核特征分析方法,提出一种对SAR目标方位角具有较强鲁棒性的非线性相关滤波器。该滤波器使用特征向量降低对训练图像的依赖性,并将特征空间进行非线性扩展,大大提高了目标的正确分类率,同时利用核函数避免了因为高维矢量而造成的计算问题。MSTAR实测SAR图像数据的实验结果表明,方法切实有效,所提出的滤波器对目标方位角失真具有较强的容忍性,不需要存储目标模板和估计目标方位角就能够实现高效率、高准确率的目标分类。

Automatic vehicle target recognition in full SAR image scenes

A new method is presented for automatic target recognition in full synthetic aperture radar(SAR) image scenes.As SAR target classification is sensitive to target’s azimuth,based on correlation filter theory and kernel feature analysis,a nonlinear correlation filter is provided,it can tolerate a distortion of target’s azimuth.The novel filter exploits eigenvectors to reduce the dependence of the training set and extends the linear combination of eigenvectors nonlinearly to improve the classification performance.Moreover,to keep the computation tractable in high dimensional space,the kernel function is employed.The tests using MSTAR database demonstrate the scheme is practical and the novel filter implements target classification efficiently and accurately without templates and target poses estimation.