摘要
提出了1种基于临界点动态调整的可扩展哈希索引算法,通过设置哈希桶容量限制,解决索引在动态非均匀数据集中表现不稳定的问题,以实现索引可扩展;通过提出1种临界点动态调整方法,解决数据点的随机偏移问题,以提高算法的稳定性。将所提算法分别在2个真实数据集和1个合成数据集上与当前主流算法进行比较。结果表明,所提算法不仅可提升检索准确率,并且具有较好的鲁棒性。
Standard Locality sensitive hashing （LSH ） is an effect ive index structure for large-scale In this paper, an improved scalable LSH for large-scale image retrieval is proposed to improve the retrieposed method , the problem of non-uniform distribution is solved by the capacity problem of near point random dritt s solved by smilarity selt-adaptation, The experimental results demonstrate that the precision and tme of mdexing of the proposed method can be significantly improved on three datasets.
作者
陈茂乾
樊皓楠
郑锦
胡海苗
CHEN Maoqian;FAN Haonan;ZHENG Jin;HU Haimiao(School ofComputer Science andEngineering, Beihang University , Beijing 100191, China)
出处
《中国科技论文》
北大核心
2017年第20期2331-2336,共6页
China Sciencepaper
基金
国家重点研发计划专项(2016YFC0801003)
关键词
局部敏感哈希
容量限制
非均匀数据集
动态调整
locality-sensitive hashing
high-dimensional feature
n on -u n i fo rm
large-scale image retrieval
