基于背景预分析的SAR图像车辆目标检测算法

Vehicle target detection algorithm based on background pre-analysis for SAR images

CFAR算法是一种常见的合成孔径雷达图像目标检测算法。CFAR算法在对背景杂波统计建模时需要考虑建模样本中是否混入目标等非背景像素。文中提出了一种新的CFAR检测算法,该算法认为目标窗口和背景窗口是同一个窗口,在CFAR算法之前,通过目标预筛选去除建模样本中的降质因素,并用广义伽马模型对背景窗口中的剩余样本建模。相比于传统的高斯模型CFAR算法,本文所用算法考虑了建模样本中降质因素对建模精度的影响,新的滑动窗模型结构更加简单,检测结果虚警率低,对距离很近的目标不会产生漏检。

CFAR algorithm is a common algorithm for synthetic aperture radar（ SAR） images target detection,needs to take whether non-background pixels are mixed in the modeling sample or not into account when modeling the background clutter. In this paper,a new CFAR detection algorithm is proposed,in which the target window and the background window are the same window. Before CFAR algorithm,the degraded factors in the modeling sample such as targets should be removed,and the generalized gamma distribution is used to model the remaining samples in the background window.Compared with the traditional Gaussian CFAR algorithm,the algorithm proposed considers the influence of the degraded factors on the modeling accuracy. The new algorithm has many advantages such as new sliding window model is simpler,the false alarm rate is low and not miss the targets which close to each other.