一种新的广义 Gamma分布下的 SAR图像目标检测算法

A Novel Target Detection Algorithm with Generalized Gamma Distribution in SAR images

摘要:

针对合成孔径雷达(SAR)图像舰船目标检测中全局广义 Gamma分布虚警数量较多的问题,文章采用了一种新的基于核密度估计(KDE)的广义 Gamma分布下的 SAR图像舰船目标检测算法。新的算法结合了 SAR图像的强度分布和空间散布分布,形成新的组合分布。在组合分布的基础上,完成 SAR图像目标检测。首先,介绍了广义 Gamma的分布模型；然后,通过核密度估计的方法估计各像素点的散布分布,结合散布分布和强度分布形成组合分布,对组合分布采用 SISE方程的方法估计广义 Gamma分布的参数,根据确定的广义 Gamma分布的表达式,求出阈值 T的解析式。通过求解逆不完全 Gamma函数得到检测阈值,进而完成舰船目标的检测。通过实测 SAR图像检测实验,对比文章算法与双参数检测算法和基于广义 Gamma分布检测算法,验证了该算法的适用性。

Abstract:

Aiming at the current situation that the number of false alarm targets is detected in overall generalized Gammadistribution in SAR images of ship target detection, this paper proposes a novel SAR images detection algorithm with gener?alized gamma distribution based on kernel density estimation. The new algorithm is a combination of intensity distributionand spatial distribution, targets of SAR images are detected on the basis of the combined distribution. First, the paper intro?duced the generalized Gamma distribution model. Then it used KDE to estimate the spatial distribution of pixels, and com?bined the intensity distribution and spatial distribution to form a combined distribution, the parameters of combined distri?bution were estimated via SISE equation. According to the formula of generalized Gamma distribution, the analytic expres?sion of the threshold T was obtained. The value of detection threshold was acquired by solving the inverse generalizedGamma function and the ship targets were detected according to the detection threshold. The performance comparisons inSAR images among double parameters detection algorithm, the detection algorithm based on generalized Gamma distribu?tion and the algorithm proposed in this paper verify the applicability of the algorithm.