数据统计查询自适应加权算法

Data statistics-based query adaptive weighted ranking algorithm

为解决汉明距离检索大量数据点共享相同汉明距离,检索结果排序模糊的问题,提出一种利用数据原始特征的加权距离检索算法。在数据集特征二值化前获得统计信息,使用查询向量的哈希特征替换二值编码计算权重值。利用数据集统计信息、查询向量和数据库二值编码计算权重值,避免由二值化引起的原始数据信息的大量丢失,更好保留了查询图像之间的差异。在两个数据集上进行实验对比,对比结果表明,该算法排序更精确,性能更优越。

A large number of data points would share the same hamming distance,which causes ambiguous retrieval results.To solve the above problem,a weighted distance retrieval algorithm based on original data features was proposed.The statistical information was obtained before the binary processing,and the hash feature of the query vector was used to replace the binary code during calculating weight value.The weighted value was calculated with the assistance of the data set statistics,query vector and binary coding,which avoided the loss of the original information and preserved the difference among the query images.Results of experiment conducted in two data sets show that the proposed algorithm has better performance.