多因素约束的自适应谐波遥感时序数据重建方法

Factors such as cloud pollution and sensor failure have greatly restricted the practical application effect of optical remote sensing time series data. At present, the data reconstruction method based on long time sequence relation only considers the dependence of single pixel time series, resulting in low accuracy of reconstruction of remote sensing time series data. Based on the harmonic analysis theory, considering the natural conditions, image spatial autocorrelation and multi-sensor auxiliary factors, we will study the adaptive harmonic reconstruction method with multiple factors to improve the reconstruction accuracy of remote sensing time series data. First, a natural condition constrained adaptive harmonic reconstruction method is proposed to mine the inherent relationship between natural conditions and remote sensing spectral information, and achieve high-precision reconstruction of high temporal but low spatial resolution data. Second, a multi-sensor spatio-temporal spectrum constrained adaptive harmonic reconstruction model is constructed with analysising the spatio-temporal correlation of multi-sensor remote sensing data and image spatial autocorrelation, and achieve high-precision reconstruction of high spatial but low temporal resolution data is. Finally, taking the Yangtze River Delta region as an example, the effectiveness of the above reconstruction method is evaluated comprehensively through the simulation experiment and the example verification. The research results of this study can improve the theory and method of remote sensing time series data reconstruction, and the generated high-accuracy remote sensing time series data has great practical value for improving the application effect of remote sensing technology.

云污染和传感器故障等因素极大地制约了光学遥感影像时序数据的实际应用效果。当前基于长时序时间关系的数据重建方法仅考虑单像元时间序列的前后依存关系，导致遥感时序数据的重建精度较低。本课题拟基于谐波分析理论，考虑自然条件、影像空间自相关性和多传感器辅助等因素，研究多因素约束的自适应谐波重建方法，提升遥感时序数据的重建精度。首先，提出一种自然条件约束的自适应谐波重建方法，挖掘自然条件与遥感光谱信息的内在关系，实现高时间低空间分辨率数据的高精度重建。其次，构建一种多传感器时空谱约束的自适应谐波重建模型，剖析多传感器遥感数据的时空谱相关关系与影像空间自相关性，实现高空间低时间分辨率数据的高精度重建。最后，以长三角区域为例，通过模拟实验和实例验证以综合评价上述重建方法的有效性。本课题的研究成果能够完善遥感时序数据重建理论与方法，同时生成的高精度遥感时序数据对提升遥感技术的应用效果具有重大的实际价值。

云污染和传感器故障极大地降低了遥感影像数 据的质量和可用性，严重制约了遥感技术的实际应用效果。因此，如何重建时序数据的缺失信息是遥感技术亟待解决的主要问题之一。本项目研究计划要点为发展时间序列数据的时域重建方法，对时间序列数据进行高精度重建，并以期在此技术支持下通过对滨海湿地分类进行验证。本研究开展了时域重建方法和滨海湿地精细分类方法的研究。并在两个方面分别取得了一定的成果。在时域重建方法方面提出了高频次遥感数据的时域重建方法、顾及物理约束的地表温度时域重建方法和基于两步融合策略的AOD数据重建方法。在滨海湿地分类方法方面提出了海岸带地物分层分类模型、基于植被物候特征的互花米草提取方法和红树林提取的时空概率阈值法，并发表了相关文章、授权了专利和获得了奖项。本研究方法对于更新长时序的遥感数据集和进行滨海湿地资源调查具有重大的意义。