融合多时空遥感数据的小水体超分辨率制图研究

Small water body is an important freshwater resource. However, due to its small size, subtle and frequent changes, it is difficult to monitor the spatiotemporal change process of small water body’s in fine-scale by single remotely sensed image. Spatial-temporal super-resolution mapping method which can produce high spatial-temporal resolution land cover maps has great potential to be used for constructing fine time-series change maps of small water bodies. But the existing methods still have shortcomings in endmember extraction and spatiotemporal model construction. To address these issues, this project will study the endmember spectra extraction methods from image sequence so as to overcome the confusion endmember problem; study the spatial distribution model of small water body from the existing high spatial resolution land cover map to solve the problems of smoothness and discontinuity caused by the traditional spatial model; construct multi-scale temporal neighborhood system and study the new model of transfer matrix which can not only inherit the detail information of sub pixels, but also keep the overall structure information of pixels; finally, propose a novel spatial-temporal super-resolution land cover mapping model to achieve the fine-scale mapping and process monitoring of small water bodies. This study will provide data support for the evolution trend of small water bodies, restoration and reconstruction of health ecological environment, and promote the theoretical research and practical application of high-resolution spatiotemporal mapping.

小水体是重要的淡水资源，但因面积小、变化细微而频繁，单一源的遥感数据难以对其时空变化过程进行高精度监测。时空超分辨率制图能生成高空间分辨率时序土地覆盖类别图，为小水体时空过程监测提供了有效手段，但现有的时空超分辨率制图在小水体端元提取和时空模型构建上仍存在不足。针对上述问题，本项目拟从小水体类别时空变化特征出发，研究时序影像上小水体端元光谱提取方法，克服小水体端元光谱易混淆的问题；通过已有高空间分辨率小水体类别图学习，研究不同形态小水体的空间分布特征，解决现有空间模式中不连续和平滑等问题；构建多尺度时间邻域系统，探讨既能继承时间维亚像元细节信息，又能保持像元级整体结构信息的地物转移概率模型；最终构建联合小水体“光谱-空间-时间”信息的超分辨率制图新模型，实现小水体精细化制图和时空过程监测，为小水体演化趋势、恢复和重建良好的生态环境提供数据支撑，推进高时空分辨率地物制图理论研究和实际应用。