基于偏振双向反射分布函数的海面溢油光学检测

Optical detection is effective for real-time monitoring and quantitative remote sensing of marine oil spill. Under the mixture of spilled oil and seawater, the difference of light intensity signal between target and background seawater decreases, which will bring uncertainties for spilled oil signal detecting. With regard to optical polarization characteristics, marine oil spill and seawater background are obviously different. But the studying on polarization mechanism of marine oil spill isn’t enough, which leads to the difficulty in analyzing its polarization characteristics in real marine environment. Therefore, the existing experimental research results about polarization cannot be directly applied into optical detection of marine oil spill. The polarized Bidirectional Reflection Distribution Function (pBRDF) is composed of BRDF and Mueller matrix, which can be regarded as the representation of target’s polarization mechanism. Mueller matrix can mathematically model complex light-matter interactions, but has difficulty performing physical modelling, which will affect the construction of pBRDF. Our research will introduce the path integral technique, which can solve the physical modeling problem of Mueller matrix. Then we will construct pBRDF of marine oil spill, and extract polarization characteristics. Polarization characteristics can be used to design detection algorithm. Finally, we will carry out the experimental verification and application research. Based on these, our research will provide reference to optical detection of marine oil spill.

光学探测技术是海面溢油实时监测与定量评估的有效手段。海面溢油与背景海水的混合，造成目标与背景的光强信号差异减小，基于图像亮度的海面溢油光学检测存在较多不确定性和难度。海面溢油与背景海水具有明显的偏振特性差异，由于海面溢油偏振机理研究的欠缺，难以描述真实海洋环境中目标的偏振特性，因此，现有的偏振实验研究结果难以直接应用到真实海洋环境中。偏振双向反射分布函数（pBRDF）由BRDF与偏振光学中的穆勒矩阵组成，是系统描述目标偏振机理的重要理论基础，但存在复杂光与物质作用过程的穆勒矩阵物理建模难题。本研究利用路径积分技术，解决海面溢油穆勒矩阵物理建模难题，研究构建海面溢油pBRDF，进一步提取目标偏振特征，拓展检测算法，开展实验验证与应用示范研究，为海面溢油光学偏振检测提供技术支撑与方法参考。

本项目首先利用模型分解、模型耦合、迭代式建模、误差机理分析、误差传播分析等多种手段，解决了复杂多自由度系统建模难题。据此，通过室内、室外实验模拟，以及各种遥感数据，建立了较完整的海面溢油pBRDF模型。基于海面溢油pBRDF模型，在特征参量的提炼过程中，不仅提取了几种常用的偏振光学特征参量，而且结合溢油表面的散射分布特性，还提炼了多种新的偏振参量——完全退偏振系数、偏振指数、平均偏振系数、偶次散射比例系数、布拉格散射比例系数、奇次散射比例系数、交叉散射比例系数等来实现海面溢油的检测。通过建立实验模型与实际遥感数据之间的数学与物理关系，明确了遥感数据中实测偏振度值与pBRDF模型及提炼的各种特征参量之间的关系。应用主成分分析、特征信号提取、神经网络等技术手段构建了一套基于偏振双向反射差异的识别算法来实现海面溢油的自动检测与分类算法。经过4 年的研究，课题组按照计划完成既定研究目标，在国内外期刊上共发表SCI检索论文20 篇，EI检索论文1篇，申请发明专利3项，授权软件著作权1项，申请软件著作权1项，培养博士3人，培养硕士5人。