多源遥感数据协同陆地辐射雾动态检测技术研究

Fog is a cloud in contact with the ground which could be dangerous for human beings, and it has serious impacts on sailing, aviation, highway transportation, military activities, etc. It also produces a great hazard to power equipment, crop growth and human health. Acoording to the limitations in the existing remote sensing fog detection methods,this proposal intends to develop a new approach for detection terrestrial radiation fog based on collaborative multi-source remote sensing imagery.The difficulty in remote sensing based fog detection lies in distinguishing cloud from fog. Firstly this proposal takes the advantage of the polar-orbiting satellite data in spatial and spectral characteristics and analyzes the difference on spectral feature between fog and cloud by simulating the spectral characteristics with the mature radiative transfer model and spectral profiles sampling from many remote sensing imagery between fog and cloud. The feature band or band combination suitable for separaing fog from cloud could be proposed, and based on this feature, comprehensive consideration of various characteristics including spectral, spatial, texture and micro-physical attributes. An object-oriented terrestrial radiation fog detection approach will be developed by using the polar-orbiting satellite remote sensing imagery. Then, according to the unique feature of terrestrial radiation fog on the movement rules and dissipation trend, it makes full use of the advantages of higher temporal resolution of geostationary satellite data. Based on the simulation of the changing spectral characteristics by the mature radiative transfer model with continuously changing observation time, and sampling spectral profiles from time series geostationary remote sensing imagery of fog and clouds, the time series feature to discriminate fog and clouds will be constructed. With this time series feature, endmembers should be extracted, identified and combined with wavelet analysis to carry out terrestrial radiation fog detection based on the time series geostationary satellite imagery. Finally, with the use of the surface weather chart datasets provided by China Meteorological Administration, these two kinds of terrestrial radiation fog detection result could be compared and analyzed. The internal relation of these two types of fog detection result will be investigated, established, and then achieve the reciprocally verification and supplement between these two kinds of detection results. The establishment of this terrestrial radiation fog detection model could collaborate polar-orbit, and geostationary satellite imagery by taking their advantages respectively, and lay a theoretical foundation and technical support for China's disaster prevention and mitigation.

雾是一种灾害性天气，对人类社会生产和日常生活都有严重影响。针对目前遥感雾检测存在精度不高，数据源单一等不足，本项目拟协同多源遥感数据开展陆地辐射雾检测研究。首先利用极轨卫星数据空间分辨率高、光谱信息丰富的优势，结合辐射传输模型，分析雾和各类云在光谱特征上的差异，构建适合云雾分离的特征波段（组合）和指标，并综合考虑云雾在光谱、空间、纹理、微物理属性等多方面的特征进行对象级雾检测研究；然后利用静止卫星数据时间分辨率高的优势，针对陆地辐射雾独特的运动和生消动态规律，在辐射模型模拟的基础上，构建云雾时序特征，通过端元选择和波谱识别，并结合小波分析进行基于时序静止卫星数据的雾检测研究；最后联合实测数据，对两类雾检测结果进行对比、统计分析，探讨其内在联系，实现两类雾检测结果的相互验证和补充，建立能够协同极轨和静止两类卫星数据各自优势的陆地辐射雾检测模型，为我国大雾减灾防灾工作奠定理论基础和技术支撑。

在本项目中（41201452），长江水利委员会长江科学院以美国EOS/MODIS卫星数据、日本MTSAT卫星数据为主要遥感数据源，通过辐射传输模型模拟云雾辐射特征和云雾样区采样统计相结合的方式，构建特征参数和时间序列特征，分别采用面向对象分类技术和云雾时序特征谱方法，开展了陆地辐射雾检测研究，利用中国气象科学数据共享服务网的全球地面天气资料定时值数据进行精度评定，发现检测精度较高，达到了预计的研究目标，为我国的陆地辐射雾检测预警奠定了良好基础，具有较好的应用前景。.在利用美国EOS/MODIS卫星数据进行陆地辐射雾检测方面，通过Streamer模型模拟各种云类和雾的辐射特征和云雾样区采样统计结合的方式，构建了归一化雾指数NDFI特征，采用均值漂移迭代合并图像分割算法，对NDFI特征进行图像分割；同时，使用盒维数法描述云类和雾的纹理差异，构建了基于NDFI特征参数加权的分形维数，用于陆地辐射雾检测，总体精度都超过了85%，kappa系数都在0.75以上，取得了较好的检测精度。.在利用日本MTSAT卫星数据进行陆地辐射雾检测方面，针对陆地辐射雾通常在晚上或凌晨形成，到第二天上午逐渐消散，有较明显的日变化特点，提出了“第一类雾”和“第二类雾”。通过Streamer模型模拟各种云类和雾的时序辐射特征，构建了云雾时序特征谱，利用非正交Haar小波基对云雾目标的热红外波段和中红外波段亮度温度差数据序列进行小波变换，并取其高频分量系数用于陆地辐射雾检测，能够比较有效的反映两种不同类型雾的持续时间和空间分布，取得了较好的检测精度。