【作者】 陈芸芝；

【导师】 励惠国； 汪小钦；

【作者基本信息】 福州大学 ， 通信与信息系统， 2011， 博士

【摘要】 水色遥感数据凭借较高的光谱分辨率和时间分辨率，以及覆盖范围广等优势，在海洋生产力评估、海洋污染监测等领域中都得到了广泛应用。在多云多雨地区，水色遥感数据应用的一个最大的限制就是数据覆盖问题，而在信息提取方面，数据的时间动态信息尚未充分应用。论文面向赤潮信息提取的应用，通过研究时间序列水色遥感图像的融合方法，进行赤潮信息的提取与预测研究。主要研究内容和结论如下：（1）多传感器水色遥感图像处理与光谱分析。对多平台水色遥感图像进行预处理，包括地理校正、大气校正、参数反演、镶嵌等过程，建立了服务于时序水色遥感图像融合和赤潮信息提取的数据集。（2）水色遥感图像融合。对叶绿素浓度、叶绿素荧光高度、443nm和667nm波段遥感反射率的数据进行比较，结果表明同一天不同传感器数据一致性良好，在空间覆盖上形成互补。提出了自适应加权平均融合算法，并把其应用于叶绿素浓度和叶绿素荧光高度融合，该融合方法能减小融合像元与邻近像元的差异，提高图像的空间覆盖率和数据采样频率，并降低图像噪声。开展GSM01模型在遥感反射率图像融合中的应用研究，该模型能充分利用光谱的冗余信息，提高融合数据的置信度，但对图像噪声敏感。（3）基于时序水色遥感图像的赤潮信息提取研究。在分析赤潮光谱特征的基础上，利用统计分析和可视化分析方法，对赤潮典型案例进行分析。发展了综合应用叶绿素浓度及差值、叶绿素荧光高度、MRI、离水反射率以及海表温度等综合特征进行赤潮信息的提取方法，并提出了简单可行的阈值确定方案。通过可视化分析方法，迅速获取参数信息在时间和空间上的变化拐点，提高赤潮监测的精度和效率。（4）赤潮信息提取技术在赤潮预报中的应用。利用融合的时序水色遥感图像和赤潮信息提取技术，以2010年福建省沿海数据为例，提取赤潮发生前期的水色异常信息，开展其在赤潮预报中的应用研究。结果表明赤潮发生前期的水色异常信息特别是叶绿素浓度和荧光高度的时空异常可作为赤潮预报的重要依据。（5）赤潮信息处理原型系统开发。基于赤潮信息提取的关键技术，构建赤潮信息处理系统，包括图像融合、赤潮光谱信息增强和时间序列数据分析等主要模块，为利用多源时序水色遥感数据进行赤潮预报提供了一个良好的平台。更多还原

【Abstract】 With high temporal and spatial resolution, wider coverage and other advantages, oceancolor data have been widely used in ocean productivity evaluation, ocean pollution, andrelated areas. But in cloud-prone and rainy area，the biggest application limitation of oceancolor remote sensing is data coverage, and the dynamic temporal information has not beenfully exploited yet. Aiming at the red tide detection, techniques of time-series ocean colordata merging were studied, as well methods for red tide extraction and change forecast wereinvestigated. The main contents are listed as following:(1) Processing and spectral analysis for multi-sensor ocean color images: multipleremote sensing images were preprocessed by geocorrecion, atmospheric correction,parameter inversion and mosaicking, then dataset for time-series ocean color datamerging and red tide extraction was established.(2) Merging of ocean color remote sensing images: comparisons among different chlaconcentration images, fluorescence line height images, remote sensing reflectance at443nm and667nm were analyzed, which demonstrate that different source data onthe same day were comparable and consistent, and can complement each other inspatial coverage. The adaptive weighted averaging method was applied on the chlaconcentration images, fluorescence line height images, which can reduce the noiseand the difference between merged and neighborhood pixels, and also can improvespatial coverage and sampling frequency. The bio-optical model of GSM01wasemployed to merge the remote sensing reflectance images and its capacity wasproved in expoiting the redundancy of spectral information and improving productconfidence,but it is sensitive to the noise.(3) Red tide information extraction based on time series ocean color images: accordingto the spectral analysis of red tide, typical red tide cases were analyzed by statisticaland visual analysis methods. Method of integrating chla concentration anddifference, fluorescence line height, modified red tide index, remote sensingreflectance and sea surface temperature was developed to extract the comprehensivered tide information, then simple and feasible threshold method was suggested.Spatial and temporal changes of red tide parameters can be achieved rapidly throughthe visual analysis method, which can highly enhance the detection accuracy andefficiency.(4) Applicaton of relevant technology in red tide forecast: using merged time series of ocean color images in Fujian coastal areas and red tide information extractiontechnology, the ocean color information especially the abnormal chla concentrationand fluorescence line height before the red tide incidents can be used as thescientific basis for prediction.(5) Development of prototype system: based on the key techniques of red tide extraction,information processing system for red tide was constructed, including imagemerging, spectral analysis, time-series data analysis and other modules, whichprovide a great platform for red tide information extraction by using time-series andmulti-sensor ocean color data.更多还原