摘要
高时间、高空间分辨率的遥感影像能够在空间、时间尺度上精细刻画植被生物物理特征和结构特性及其在空间、时间上的变化,对监测植被生态特征有着重要的作用.为有效记录地表特征的变化,本文提出了一种基于像元降尺度的时空遥感数据融合算法(downscaling difference spatial and temporal data fusion algorithm,DDSTDFA).该方法分别开展基于Landsat、MODIS和NOAA影像的模拟与真实实验,与已有STDFA(spatial and temporal data fusion algorithm)和FSDAF(flexible spatiotemporal data fusion)进行对比.结果表明,DDSTDFA算法表现出以下优势:1)DDSTDFA算法能够同时预测地表特征发生的多种变化方向,改进了基于像元分解算法的缺陷,与STDFA相比在变化区域表现出更高的精度;2)DDSTDFA融合影像的空间分布特征更接近真实影像,消除像元分解融合法中常见的"图斑""边界"问题;3)与FSDAF算法相比,DDSTDFA算法在保证影像融合精度的前提下,运行速度提高了50%~60%.因此,DDSTDFA算法更适合于大范围高时空影像数据融合,为地表动态监测提供丰富的遥感影像数据源.
High spatiotemporal remote sensing data can be used to describe biophysical and structural characteristics of vegetation and phonological changes, playing an important role in vegetation monitoring. The aim of the present work was to improve fusion-data performance in areas where land surface characteristics changes occur in different directions and to propose an easy and efficient method for spatiotemporal data fusion： Downscaling Difference Spatial and Temporal Data Fusion Algorithm （DDSTDFA）, to fuse Landsat 8 OLI fine-resolution images and MODIS/NOAA coarse-resolution images. This method was compared with STDFA and FSDAF. DDSTDFA algorithm can predict changes in surface characteristics occurring in different directions simultaneously, and improve defects of un-mixing based algorithms. Compared with STDFA, DDSTDFA showed higher accuracies for different areas with changing land-surface characteristics. DDSTDFA- predicted images resembled real images by visual analysis. Block and patch effects found in previous methods were effectively avoided. Compared with FSDAF algorithm, efficiency of DDSTDFA was improved hy 50-60% with high accuracy. DDSTDFA is therefore more suitable for wide-range high spatiotemporal image datafusion, this will provide a wealth of remote-sensing image data sources for land surface dynamic monitoring.
出处
《北京师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2017年第6期727-734,共8页
Journal of Beijing Normal University(Natural Science)
基金
国家重点研发计划"粮食丰产增效科技创新专项子课题"资助项目(2017YFD0300402-6)
