多源土地覆被遥感信息融合及数据重构研究

发布时间：2018-09-17 19:04

【摘要】：针对当前土地覆被遥感数据存在的问题及现有融合算法的不足提出了基于Dempster-Shafer理论和贝叶斯最大熵（Bayesian Maximum Entropy，BME）理论的多源土地覆被遥感数据融合及重构框架。研究首先评价了四种土地覆被遥感数据在国家及区域尺度的类别精度及混淆特征，构建了多源数据的空间一致性及类别均质性图谱。研究提出了基于LCCS分类体系和最小距离向量计算土地覆被遥感数据及MODIS LAI针对目标分类体系的基本概率函数的计算方法，，实现了Dempster-Shafer理论框架下多源土地覆被遥感数据及辅助信息融合，并从多角度对融合结果进行了评价。针对典型研究区域，对比分析了指示克里金与贝叶斯最大熵模型在宏观及微观尺度的数据重构结果，挖掘了预测位置邻域内已知点个数对于预测精度及预测时间的定量影响，分析了在软数据精度较低情况下贝叶斯最大熵模型联合使用硬数据及软数据相对于单独使用硬数据进行数据重构的结果。
[Abstract]:Aiming at the existing problems of land cover remote sensing data and the deficiency of existing fusion algorithms, a framework of multi-source land cover remote sensing data fusion and reconstruction based on Dempster-Shafer theory and Bayesian maximum entropy (Bayesian Maximum Entropy,BME theory is proposed. Firstly, the classification accuracy and confusion characteristics of four land cover remote sensing data at national and regional scale were evaluated, and the spatial consistency and class homogeneity map of multi-source data were constructed. Based on LCCS classification system and minimum distance vector, the method of calculating land cover remote sensing data and the basic probability function of MODIS LAI for target classification system is proposed. The multi-source land cover remote sensing data and auxiliary information fusion are realized under the framework of Dempster-Shafer theory, and the fusion results are evaluated from many angles. In this paper, the data reconstruction results of Kriging and Bayesian maximum entropy models at macro and micro scales are compared and analyzed for typical study areas, and the quantitative effects of the number of known points in the prediction location neighborhood on the prediction accuracy and prediction time are excavated. The results of using hard data in Bayesian maximum entropy model and reconstructing soft data in comparison with hard data alone are analyzed in the case of low precision of soft data.
【学位授予单位】：中国矿业大学（北京）
【学位级别】：博士
【学位授予年份】：2013
【分类号】：P237

【参考文献】