空间数据压缩的高光谱降维技术比较

摘要

With the rapid development of hyperspectral remote sensing,the inefficiently retrieved information from the massive remote sensing images with amount of bands has become a problem to be urgent solved.In this paper,the curve data compression algorithm is introduced in the spectral data dimension reduction.Based on rule tree grouping and curve based data integration technology,it designed the 16 fork tree transformation,the offset spectroscopy retrieval,angle retrieval,and douglaspoke spectral search operator.It used three statistical methods which are relative spectral discriminatory probability,relative spectral discriminatory power,and relative spectral discriminatory entropy in contrastive analysis from different aspects.Compared with the conventional SAM and SID,the new retrieval operators reduced the time and frequency spectrum retrieval.In order to ensure the condition of similar recognition ability,the new operators greatly improved the retrieval efficiency of the program,which proves to be the fast and efficient hyperspectral feature matching and retrieval methods.%针对从波段数目较多的海量高光谱遥感影像数据中高效地检索出所需信息这一迫切需要解决的问题,将空间数据压缩算法引入到光谱数据降维中,以规则树分组和曲线数据综合技术为基础,设计了十六叉树状变换、垂距光谱检索、偏角光谱检索、道格拉斯普克光谱检索算子.采用相对光谱识别概率、相对光谱识别熵、相对光谱识别力3种统计学方法,从不同角度通过与常规的光谱角度制图和光谱信息熵进行比较分析发现:利用新的检索算子提取光谱曲线特征向量,进行相似性测度,降低了光谱检索的时间频率.在保证相近识别能力的条件下,能够大大提供高程序的检索效率,是几种快速有效的高光谱特征匹配和检索算子.