摘要
遥感时空信息融合需要其他时相的数据作为辅助,而现有的时空融合应用往往随机选用单时相或双时相辅助数据,从而导致融合结果并非最佳。单/双时相辅助数据中,哪种能获得更优的融合结果?如何选用单/双时相辅助数据进行最佳融合?这是目前研究中尚未探究的问题。针对此问题,该文对单/双时相辅助数据的融合结果进行了系统比较,总结了单/双时相辅助数据在不同时相变化特征下的适用规律。结果发现,当存在多个可供选择的辅助时相时,若预测时段内时相变化幅度基本均等,应优先选用双时相辅助数据进行融合;若时相变化幅度明显不等,应优先选用差异较小的单时相辅助数据进行融合。此外,为深入揭示时相差异与最优辅助时相的关系,该文对其进行了量化研究,提出了一种用于选取最优辅助时相的经验性方法。
Observations on reference dates, serving as auxiliary data, are necessary in spatiotemporal fusion. Existing applications with spatiotemporal fusion tend not to take the temporal characteristics of data into acc：ount, and their random selection of auxil-iary data often results in the decline of prediction accuracy. Which one can acquire better predictions between mono- and b- tem-poral auxiliary fusion and how to choose the optimized auxiliary data？ These problems need to be solved.This paper presents a detailed comparison between predictions with mono- and b- temporal auxiliary data, and concludes their adaptation rules under different situations of temporal change. Results indicate that, when temporal change of prediction periods is consistent, b- tempo-ral auxiliary data should be preferred. In contrast, mono- temporal auxiliary data is a priority wThen sudden change happens in one period. Besides, in order to quantitatively describe the relationship mentioned above, an eimpirical method for selecting optimized auxiliary data is proposed. With the index Rtc and corresponding threshold range, the optimized reference data can be selected and it w ill contribute to the prediction accuracy to some extent.
出处
《地理与地理信息科学》
CSCD
北大核心
2017年第5期9-15,共7页
Geography and Geo-Information Science
基金
国家自然科学基金项目(41422108、41601357)
湖北省自然科学基金创新群体项目(2015CFA002)
中国科学院创新交叉团队项目
