一种基于全极化SAR数据广义多子视相干的海面船只目标检测方法

摘要

The traditional ship detection algorithms,such as 2L-IHP (Two Looks Internal Hermitian Product),Pol-IHP (Polarimetric Internal Hermitian Product),etc,usually utilized two sub-look images cross-correlation to decrease omit-ted detection for small ships.However,because they were constrained by the number of sub-look images,the previous meth-ods could not increase ship-sea contrast to much extent,which affected the ship detection accuracy.Therefore,this paper pro-poses a detection algorithm based on generalized multi-sublooks correlation using polarimetric SAR (POLSAR)data.First-ly,the sub-look decomposition method is applied for POLSAR data to get multi-sublook POLSAR images.Then the correla-tion matrix and the coherence operator based on the generalized similarity parameter (GSP)are defined to calculate the co-herence image of the multi-sublook images.Finally,the constant false alarm rate (CFAR)detection method is utilized for ship detection by the calculated cumulative distribution function (CDF)of the coherence image.The experiments prove that ship-sea contrast can be increased with the number of sublook images by the propose method,which reduces the undetected probability of the ships and also improve the ship detection accuracy significantly.%传统基于子视相干的检测算法，如2L-IHP（Two Looks Internal Hermitian Product）和Pol-IHP（Polarimetric Internal Hermitian Product）等，通常利用两个子视影像进行相干来检测海面弱小船只目标。但受子视影像个数的限制，无法大幅度地提高船海对比度，进而影响了检测精度。针对该问题，本文提出了一种基于全极化SAR的广义多子视相干检测算法，首先利用子视分解方法对全极化SAR数据进行处理得到多个子视全极化影像；接着，基于广义相似性参数（Generalized Similarity Parameter，GSP）定义这些子视影像间的相关矩阵和相干算子来计算相干图；然后，利用恒虚警率（Constant False Alarm Rate，CFAR）检测方法结合统计的相干图累积分布函数进行船只目标检测。通过实验，表明利用本文算法船海对比度随着子视影像个数的增加而得到大幅提高，从而减少了弱小船只目标的漏检，显著提高了船只目标检测精度。