张量框架下高维遥感影像空-谱协同解译方法研究

The developments in earth observation techniques improve the spatial, spectral, and temporal resolution of remotely sensed imagery. As the image resolution evolves, data dimension and size substantially increase, and the spectral features from the same class become more diverse and heterogeneous, which determines that spectral information alone is not adequate for accurate feature presentation and obtaining high-accuracy geospatial information. Consequently, it can be regarded as the core scientific problem of high performance interpretation for remotely sensed imagery to develop the collaborative exploitation for spatial and spectral information in remote sensing imagery while considering dimensional structure. Presenting and analyzing the remotely sensed imagery in tensorial form contribute to preserving the dimension relationships and feature structure. Accordingly, in this project, we propose a series of innovative models to solve the aforementioned scientific topic by introducing tensor algebra, including: tensor feature extraction, support tensor machine classification framework, and tensor feature based change detection. The proposed framework will be validated with data collected from two research regions covering Beijing city and Wuhan city. This research is potential for improving the intelligent processing level of remotely sensed data and is expected to exceed the current technological bottlenecks. It has important academic value and practical significance in remote sensing.

对地观测技术的发展提升了遥感数据的空间、光谱和时相分辨率，此时纯光谱信息已不足以满足影像精确解译的要求。协同解译空间、光谱信息，是解决当下海量遥感数据高效解译的关键科学问题。如何在解译过程中顾及空-谱结构关系，发展高维遥感数据的协同表达与智能化理解，是本课题的研究重点。由于张量能够直观表达遥感数据的空间、光谱信息，张量学习可以完整地保持数据的高维结构关系，本课题将张量表达引入遥感数据处理，围绕这一科学问题展开研究，分别提出张量特征提取模型、支持张量机分类、融合空-谱变化的张量变化检测框架等一系列原创性的方法，并拟针对北京和武汉两个实验区进行方法验证。研究成果能显著提升当前高维遥感数据的智能化处理能力和信息利用程度，有望突破当前的解译方法无法充分利用空谱信息的技术瓶颈，具有重要的学术价值和现实意义。

协同解译空间、光谱信息，是解决当下海量遥感数据高效解译的关键科学问题。本课题的研究重点是在解译过程中顾及空-谱结构关系，发展高维遥感数据的协同表达与智能化理解，其主要成果包括：顾及维度相关性的张量特征提取、面向高维数据的支持张量机分类框架、基于张量特征的变化检测框架等原创性方法，以及对张量分解的地学语义挖掘的应用探索。在北京、武汉实验区的真实遥感数据实验表明，本课题发展的基于张量的空谱协同处理框架，顾及了“高维、异构”的数据特点，能够保持数据的空间、光谱多维结构关系，克服现有基于向量算法在处理流程上的不足，提升遥感数据智能化解译的应用潜力。其研究成果为高维遥感数据的智能化处理能力和信息利用提供了新的思路和依据。项目资助发表SCI论文一篇，待发表两篇。项目投入经费14万元，支出9.15万元，各项支出基本与预算相符。剩余经费4.85万元，计划用于本项目后续研究支出。