基于一种新型去雾降噪算法的目标偏振信息获取技术及其在透雾成像中的应用研究

The polarimetric optical imaging technique is a novel imaging method which can obtain the polarization information of the light reflected by the objects. It can show completely different and unique information compared to the intensity and spectrum information. So, this technique has many potential application values in military and civil fields. However, in the hazy condition, it will be difficult to precisely obtain the polarization information reflected by objects, because the scattering effect will not only reduce the polarization information of the light reflected by objects, but also produce stray light with partial-polarization information. This drawback confines the applications of polarimetric imaging technique. We aim to systematically study polarization-information retrieval technology of targets and its application in hazy conditions. The main contents are as follows: to study the polarization properties of the stray light scattered by the haze, to reveal the natural differences between the reflected light and the stray light, as well, to build a new method used to estimate the stray light fast and precisely; Based on this research, we could establish a new polarimetric dehazing algorithm which can achieve single-image dehazing; Finally, the image that with the polarization characteristics of objects (e.g. degree of polarization or angle of polarization) and the corresponding experiments can be further studied. This project has important application values and scientific meanings in improving the applicability of this polarimetric optical imaging technique in hazy condition.

偏振光学成像技术是一种基于对目标反射光偏振信息获取的新型成像技术，可以使目标呈现出与强度信息和光谱信息完全不同的独特信息，在军用和民用领域均表现出非常重要的应用价值。然而在雾霾条件下，雾霾颗粒的散射作用不仅会削弱目标反射光的偏振信息，而且会产生具有一定偏振信息的大气散射光，从而给目标反射光偏振信息的准确获取带来了困难，限制了偏振光学成像技术的应用。本项目拟开展雾霾气象条件下的目标偏振信息获取技术及其透雾成像应用研究。具体研究内容：研究雾霾颗粒对大气光的散射偏振特性，揭示目标反射光与大气散射光偏振特性的本质区别，设计一种精确估算大气杂散光强的新方法；基于此方法建立一种可直接对单幅图像进行去雾的偏振成像降噪算法；基于该算法开展目标偏振度、偏振角等偏振图像信息的高保真获取技术及透雾成像实验应用研究。本项目的开展对提高偏振光学成像技术在雾霾等不良气象环境中的适用性具有重要的应用价值和科学意义。

本项目的研究目标为提出一种可以在雾霾天气条件下使用的目标偏振信息精确获取技术。项目执行期间完成了以下研究内容：1) 对现行偏振成像去雾技术的特点与缺陷进行了深入研究，提出了采用邻域滤波降噪方法减小量子噪声对偏振去雾结果的影响，提高了偏振去雾技术在浓雾霾天气条件下的去雾能力。进一步，将优化后的偏振去雾技术与红外去雾技术相结合，提出了可见光与近红外融合的偏振光学成像去雾技术。2) 对大气光强的估算精确性和实时性直接影响偏振成像的效果和效率。提出了一种结合暗通道先验原理自动判定天空区域的偏振去雾算法，自动选择图像中的天空区域以精确估算大气光强，实现了偏振去雾处理的自动化。进一步，采用HSI颜色空间转换，通过单通道偏振去雾处理实现彩色图像去雾和颜色恢复，偏振去雾处理效率提高了3倍。3) 与实际应用相结合，开展了全偏振分孔径偏振成像系统研究。偏振成像相机通过孔径分光的方式将入射光一分为四，通过不同偏振片的调制后分别聚焦到单探测器的四个象限中，实现单镜头单探测器的全偏振信息采集。利用该偏振成像系统开展了实时偏振去雾技术研究、水下偏振成像去散射技术研究和全偏振穆勒矩阵测量方法研究，为偏振成像技术的实用化奠定了理论基础和硬件设备。