太阳光入射海表特定区域形成海洋耀光，呈强反射和偏振特性。在卫星海洋遥感中，海洋耀光对遥感成像质量有较大影响，尤其对海洋上空云、气溶胶及海色等研究干扰较大，因此剔除海洋耀光是卫星遥感数据处理过程中首先要解决的重要问题。以PARASOL卫星的POLDER3载荷数据为研究对象，获取载荷成像时刻太阳及观测几何、海表风速风向及气溶胶光学厚度等参数，采用海气辐射传输相关理论，结合耀光角和多角度NIR偏振辐射信息（865 nm），构建基于近红外偏振辐射特性规则的OGDD模型，获取耀光角临界值实现耀光动态检测。以印度洋某海区为研究对象，获取实测耀光区NIR通道的大气层顶反射率和偏振反射率，利用OGDD模型将耀光角判别临界值调整为34°。相比MODIS经验临界值（40°），标记耀光像元相对减少了30%。结果表明，该方法不仅能通过动态调整临界值准确识别海洋耀光，为云检测及云物理特性反演提供可靠的源数据，还能为高分五号卫星多角度偏振载荷在轨定标及气溶胶反演提供支持。

Sunlight incoming sea surface forms ocean glint (OG) in a special area and shows strong reflection and polarization characteristics. OG significantly influences the imaging quality of ocean remote sensing, especially because the interference is large for clouds and aerosols above the ocean surface. Therefore, eliminating OG is the key problem to be solved in the process of remote sensing data. At present, most OG detections of sensors are utilized by a rough sea surface polarization model combined with an empirical threshold. Orbit height and resolution is different from various sensors available. Thus, the result is quite inaccurate using the same threshold to discriminate glint pixels, thereby resulting in several pixels being utilized ineffectively. The Chinese GF-5 satellite has been scheduled for launch in 2017. It carries a Directional Polarimetric Camera (DPC) sensor for the atmospheric polarization research at a global scale. Similarly, the OG detection is essential to DPC data processing. The traditional method cannot achieve the dynamic detection of OG for different satellite data. Thus, the problem of accurately obtaining a threshold angle has become the key difficulty of OG dynamic detecting research.
OG dynamic detecting (OGDD) method was proposed on the basis of near-infrared (NIR) polarized data, which were not easily affected by atmospheric disturbance under a clear sky. From these data, several parameters, such as the geometry conditions of solar and observation and sea surface wind speed and direction, could be obtained. According to ocean–atmosphere coupled radiative transfer theory, the OGDD model, combined with the OG and multi-directional NIR-polarized radiation information (865 nm), was developed on the basis of the regular performance of the NIR-polarized radiation characteristic. The OGDD was realized by obtaining the dynamic threshold of a glint angle. A calibration layer was selected by using the polarized characteristic tendency of an OG center from detecting layers, and after cloudy pixels have been removed, a slope dynamic analysis was conducted on the basis of an OG-polarized radiative regular variety on top of the atmosphere. Finally, the ocean pixels were marked as glint pixels using the dynamic threshold of the glint angle.
This study used PARASOL/POLDER3 satellite data as the research object for the DPC simulation and selected the Indian Ocean as the study area. The NIR channel utilized the OGDD model to acquire apparent and polarization reflectivity (865 nm) of the sea surface. The glint threshold was adjusted dynamically to 34° using the OGDD model. In comparison with MODIS 40°, the glint angle was reduced by approximately 15%, and the pixel-marked glint was relatively decreased by 30%. Similarly, in comparison with POLDER3 30°, the glint angle was relatively improved by nearly 13%, and the pixel-marked glint was increased by 30%. The model can effectively distinguish the glint and non-glint pixels through a dynamically adjusted threshold and significantly improve the utilization rate of the pixels by reducing the interference of aerosol retrieval at the clear sky area. Furthermore, this model can generate reliable data for cloud detection and microphysical characteristic retrieval and provide support to developing the in-flight calibration and aerosol inversion of the GF-5 satellite multi-angle polarization sensor.