基于稀疏转换学习的遥感影像时空融合模型与方法研究

现有在轨卫星遥感传感器不能满足日益深入的遥感应用对时间分辨率和空间分辨率综合精准度和充分性的要求。本项目旨在发展一种基于稀疏转换学习的遥感影像时空融合模型，来虚拟生成高空间与高时间分辨率遥感影像。它的主要特点是提出在影像隐含的稀疏编码空间中来建立高/低分辨率影像间、以及影像反射率随时相变化的对应转换关系，从而实现遥感数据在空间维的增强和时间维的扩展。项目将以“影像稀疏转换学习模型—遥感影像¬的时空融合方法—遥感影像融合效果的验证”为研究主线，围绕着遥感影像的预处理、影像稀疏转换模型的建立与关键算法、模型的结构化表达与非线性扩展、不同条件下的影像时空融合方法、以及模型的验证与应用等关键内容开展系统而深入的研究。该研究成果能够从多源多时相的遥感数据中综合出可应用的高空间、高时间分辨率的遥感数据，使其满足遥感应用在时-空-谱等方面的综合要求，具有重要的理论与实用价值。

The current operating satellite sensors can not meet the reqiurements of various in-depth remorely sensed applications on high-temporal, high-spatial and high-spectral resolutions. This proposal proposes a novel solution on spatiotemporal scale fusion of remotely sensed image based on structured sparse transfer representation model to synthesize the high-temporal and high-spatial resolution images from multi-sensors and multi-temporal available images.The main feature of our model is that the transferring mapping functions between higher and lower images, as well as the multi-temporal images are explored and then built in the hidden sparsing coding space.The main research context of the proposal is mainly focus on image preprocessing, sparse transfer representation model and corresponding algorithms,structured representation and nonlinear expension model, and spatiotemporal scale fusion of remotely sensed data with different situations, as well as validations on the model and algorithms with multi-resource images acquired from different sensors.The project will deliver applicable high-spatial and high-temporal resolution satellite images that are improtant data sources for environmrntal change modeling,monitoring and analysis, thereby overcoming the bottlenecks of current earth obervations. The resultant images also lay good foundation for subsequent image processing, and hold great value to various remote sensing applications required high spatio-temporal-spectral resolution image data.