基于改进PT的遥感高光谱相似性查询优化算法

Remote sensing hyperspectral similarity query by using improved pyramid technique

针对传统金字塔技术(PT)不能直接处理相似性查询的不足,提出了一种改进的PT索引结构,该索引对传统PT索引机制进行改进,在降维映射函数中以数据点到中心参考点的距离值代替传统PT索引中的特征维度高度值,使改进后的PT索引具有距离查询处理能力,并设计了完整的相似性查询处理方法.以06版USGS典型地物光谱库为实验数据集,与传统PT索引相比,改进后索引的查询处理总时间平均缩短了1/3,而且对维度变化更加不敏感.

The similarity query can not be processed directly by pyramid technique （PT）. An improved indexing structure was proposed. This indexing can utilize the distance between the data point and the central reference point instead of the height of characteristic dimensionality in the mapping function of dimensionality reduction. Through the improvement of the indexing mechanism, the new indexing provides the capability to directly process the similarity query based on the spatial distance. The complete method to process the similarity queries was designed by this improved indexing, and the 06＇s specific earth object spectral library from USGS was taken as the experimental dataset. The results of the experiments demonstrate that compared with the traditional PT the overall query processing time of the improved indexing decreased about 1/3, and it also has less sensitiveness to the constant increasing of dimensionality.