为有效利用简缩极化SAR进行海洋溢油检测，本文基于简缩极化特征值分析，提出了3个用于简缩极化溢油检测的参数，引入了基于简缩极化特征值分解的简缩极化熵Hc（Compact Polarization Entropy）、简缩极化比参数PFc（Compact Polarization Fraction）、简缩极化基准高度PHc（Compact Polarization Pedestal Height）特征进行海洋溢油检测。海表的散射类型主要为低熵散射（小粗糙面发生的Bragg散射），为弱去极化、弱散射过程随机性状态，由于溢油会阻尼海水的Bragg散射，使其熵值变高、呈去极化、强散射过程随机性状态，故简缩极化熵、简缩极化比参数和简缩极化基准高度可以用来检测海洋溢油。本文采用C波段的Radarsat-2、SIR-C/X-SAR数据进行了实验，结果表明：简缩极化熵、简缩极化比参数和简缩极化基准高度能够有效抑制疑似溢油，使海水与疑似溢油差异变小；突出溢油区域，使海水与溢油的可区分性变大。

Oil spill is one of the most common causes of marine pollution. Synthetic aperture radar (SAR) can detect marine oil spills due to its all-weather and all-day imaging capability. However, conventional algorithms using single-channel SAR usually confuses oil spills with oil spill lookalikes, making oil spill detection a challenging task. In recent years, polarimetric SAR has been studied to improve oil spill detection performance, particularly with respect to distinguishing oil spills and lookalikes. Compact polarimetric (CP) SAR is one of the most popular trends of polarimetric SAR platforms. Compared with full polarimetric SAR, CP SAR contains abundant polarization information of scattering targets. CP SAR also has large imaging range and its manufacturing cost is considerably low. Therefore, CP SAR is a promising tool for oil spill detection.
In this study, we utilized CP HP-mode data and derived three parameters to distinguish oil spills from lookalikes. The three novel parameters are CP entropy (Hc), CP fraction (PFc), and CP pedestal height (PHc), which are conducted from CP eigenvalue decomposition. The scattering type of sea surface is dominated by low entropy scattering. For instance, Bragg scattering from small rough surfaces can be regarded as weakly depolarized. In this vein, CP eigenvalue ? >> 1 is much higher than ? >> 2, which indicates the dominance of Bragg scattering. On the contrary, oil spill weakens the Bragg scattering of the ocean, and the difference between the two eigenvalues is not as high as before. Based on former consideration, Hc is high, PFc is low, and PHc is high, which means that the average scattering is in a depolarized state. The target is considered "random", and polarization status information is lost. Hence, the aforementioned parameters can be used to detect oil spills. The threshold is determined by Otsu's method and statistical analysis. The threshold for PFc and PHc was automatically determined by Otsu's method, and other parameters were statistically analyzed to determine the threshold.
In this study, Radarsat-2 and SIR-C/X-SAR data of C-band were used to detect oil spills. The results show that Hc, PFc, and PHc can effectively distinguish oil spills and lookalikes. The performance of all the parameters based on the CP eigenvalue decomposition is excellent in effectively excluding lookalikes, highlighting oil spill areas, and enhancing differences between oil spill and sea clutter. Moreover, the parameters are excellent in effectively maintaining details.
CP SAR is promising in polarimetric SAR system because of its ability to detect a wide range of marine oil spills. The parameters derived in this study, which were based on the CP eigenvalue decomposition, can obtain satisfactory results in distinguishing oil spills from lookalikes. The quantitative experiments also confirm that the Bhattacharyya distance between oil spills and sea surface surpasses that of lookalikes, which proves that our method achieves fine oil spill detection.