基于Lucene的图像检索实验设计

The Design of Lucene-based Image Retrieval Experiment

从海量图像数据库中快速有效地查询和检索所需要的图像已经成为一个研究热点,学者们对图像检索技术进行了深入研究。目前,广泛应用的是基于内容的图像检索技术。本文研究了基于Lucene的图像检索实验的设计。介绍了基于Java的全文检索引擎Lucene;对在Lucene基础上开发的LIRe架构了进行了分析,提出了基于LIRe架构的可扩展图像检索实验方案;最后,以时尚品图像检索为应用案例,遵循该方案搭建实验环境,实现了时尚品图像检索原型系统,并用查全率和查准率作为评价准则对实验结果进行了分析。实验案例表明,本文实验设计具有很好的可操作性和可扩展性。

It has become a hot research topie to search and retrieve image from massive database, and scholars have carried out deep research on image retrieval technology. Nowadays, content-based image retrieval teehnology has been widely used. This paper researches the design of Lucene-based image retrieval experiment. Firstly, this paper introduces Java-based full-text search engine -- Lueene; then, analyses the architecture of LIRe which is developed on the foundation of Lu~＇ene, and l^UtS forward LIRe-based sealable image retrieval scheme; finally, uses fashion gomls image as an example, builds up experiment environment followed hy the scheme, carries out fashion goods image retrieval prototype system, and analyses the results of the experiment by a criterion of recall ratio and preeision ratio. Experiments show that the experiment design has good maneuverability and scalability.