空间信息网络下的高光谱遥感协同观测理论与方法研究

This proposal will fully take advantages of hyperspectral remote sensing (HRS) which is the integration with spatial information and spectral information, and is suitable for automatically processing. The key role of HRS in the spatial information network will be efficiently played to satisfy the applications on urban classification, target detection, and quantitative inversion of vegetation and water parameters. We will establish sparse features representation model of HRS, and develop high efficiently synergistic observation modes and technique flows of integrating hyperspectral sensor with visible and infrared sensors. Then, based on the overall consideration of hyperspectral data and high spatial data, emitted radiation data, we will pertinently propose the algorithms and models of image positioning and geometric correction, fusion processing and typical applications. Based on full exploration of inherent high dimensional sparse features of HRS and parallel computing characteristics of synergistic observation data, we will develop the algorithms for onboard high precision real-time geometric positioning and multi-mode real-time processing for thematic information extraction. Fully considering the particularity of the processing on the satellite, we will introduce high performance and low power parallel computing hardware, to develop a prototype with the integration of hardware and software for data rapid processing. Comparing to technical procedures used in traditional remote sensing data processing on satellite, the processing efficiency of our developed techniques in the prototype will achieve more than 2 times. These research achievements will provide theories and methods support for the synergistic observation and comprehensive applications of integrating hyperspectral sensors with and other sensors in the framework of the spatial information network.

本项目将充分利用高光谱遥感图谱合一以及适合自动化处理的优势，高效发挥其在空间信息网络中的关键作用，面向城镇地物分类、目标探测、植被和水体参数定量反演等典型应用需求，建立高光谱数据稀疏特征表达模型，提出高光谱传感器与可见光、红外等传感器的高效率协同观测模式和技术流程，有针对性地发展综合高光谱数据和高空间数据、发射辐射数据等多源异质数据的图像定位与几何校正、融合处理和典型应用的算法模型。充分挖掘高光谱数据内在的高维稀疏特征以及协同观测数据的并行计算特性，提出星上高精度实时几何定位以及面向星上专题信息提取的多模式实时处理算法，并针对星上处理的特殊性，引入高性能低功耗并行计算硬件，搭建软硬件一体化快速处理原理样机，相对于传统星上遥感数据处理技术流程实现处理效率2倍以上的提升。本项目研究将为实现在空间信息网络框架下的高光谱与其它多源传感器协同观测和集成应用提供理论和方法支撑。

高光谱以其图谱合一的独特优势在空间信息网络中有着不可替代的作用。本研究旨在高效发挥高光谱在空间信息网络中的作用，实现高光谱与其它多源传感器的无缝集成和高效协同，提升空间信息网络中遥感信息获取与应用的水平。当前，制约空间信息网络下的高光谱协同观测的瓶颈问题主要三个：多源异构平台的高效协同观测共性处理技术；多源异质遥感图像特征表达模型与融合应用；高维稀疏数据星上快速处理技术。随着遥感技术的快速发展，特别是高光谱卫星的业务化运行，势必为空间信息网络建设带来新的机遇和挑战，迫切需要针对上面三个问题进行理论探索和技术攻关，力争最大限度发挥高光谱协同观测的作用。.项目成果支持下，获得国家科技进步二等奖1项、测绘科技进步二等奖1项、国家基金委优青人才2项、北京市杰出青年人才1项、北京市科技新星计划1项等，晋升IEEE Fellow 1名，举办第五届全国成像光谱对地观测学术研讨会，先后在IGARSS、WHISPERS、全国成像光谱大会等学术国内外会议交流30余次，产生了一系列国内外学术影响。项目资助发表论文83篇（其中，在IEEETGRS、ISPRS、JRTIP等主流SCI期刊发表63篇)，获得国家技术发明专利6项，培养研究生36名，唯一资助出版《高光谱图像协同处理理论与方法》，共同资助出版《岩矿高光谱遥感》，研究成果应用于国产高分卫星数据的处理，直接服务于上海卫星工程研究所、国家自然资源督察济南局以及上海微小卫星工程中心等业务运行需要。