基于MapReduce的外观设计专利图像检索方法  

Retrieval of Design Patent Images Based on MapReduce Model

作　　者：王贤伟[1] 戴青云[1] 姜文超[1] 曹江中[1] 

机构地区：[1]广东工业大学信息工程学院,广州510006

出　　处：《小型微型计算机系统》2012年第3期626-632,共7页Journal of Chinese Computer Systems

基　　金：2008广东省省部产学研结合项目(2008B0900254)资助;2008广东省现代信息服务业重点项目(GDIID2008IS005)资助;2008广东省现代信息服务业发展专项资金扶持项目(GDIID2008IS007)资助

摘　　要：工业设计领域中,对外观设计知识产权保护的需求日益增大,为解决外观设计专利图像检索数据的快速增长与检索实时性之间的矛盾,将基于内容的图像检索技术应用于MapReduce框架中,开发了外观设计专利智能检索系统.Map任务提取示例图像的形状、纹理以及颜色特征,与专利特征库中的特征进行相似度匹配计算,Reduce任务对Map任务的匹配结果进行相似度排序,得到检索结果.实验表明,该方法能够均衡系统负载,提高资源利用率,有效降低了在大数据集上进行图像检索的时间.In the field of industrial design,the demand for intellectual property protection has been increasing.To settle the conflict between the rapid growth of the data which is in the database of patent design image retrieval and the demand for real-time retrieval,we combine the content-based image retrieval technology with MapReduce framework,develop the intellectual content-based image retrieval system.The mission of Map is to extract the feature of shape,texture and color in the retrieval image,which is used to match with the features in the database of patent feature.The mission of Reduce is to sort the matching result according to the similarity,and finally get the retrieval result.The result of our experiment shows that the algorithm we proposed can not only balance the load of system,but also can increase the utilization ratio of the resources,which significantly decrease the retrieval time in the large data sets.

关 键 词：MAPREDUCE 外观设计专利 图像检索 分布式计算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]