Design And Implementation Of Distributed Design Patent Retrieval System Based On Hadoop

In recent years,with the increasing awareness of intellectual property protection of innovation subjects,the amount of design patent data has increased rapidly,and the design patent image retrieval system in a stand-alone environment has become increasingly difficult to meet the needs of users for fast retrieval;and distributed storage and distributed computing models can be applied efficient storage and retrieval of design patent.In order to cope with the challenges brought by the rapid increase in the amount of design patent data,in view of in a stand-alone operating environment,the design patent retrieval system is inefficient and poor in scalability.This article has made an in-depth study and research about the content-based image retrieval technology and distributed technology.A Hadoop-based distributed image retrieval system for design patent is proposed,which implements design patent retrieval on a distributed platform,and optimizes the process of indexing and retrieval.The main research contents are as follows:1)Analyzed the development status and problems of image retrieval technology and Hadoop platform,and expound the feasibility and necessity of distributed design patent retrieval platform.The key technologies of content-based image retrieval,full-text retrieval technology,and distributed computing platform Hadoop are researched respectively,and the full-text retrieval technology is used to establish inverted index and retrieve image data;the two major modules of Hadoop--distributed File system and distributed computing framework Map Reduce,are used to store and retrieve design patent data.2)Designed a distributed appearance patent search system based on Hadoopd.Used the distributed file system of the Hadoop platform and the Map Reduce framework model,combined with full-text search and CBIR technology,parallelized the establishment of an inverted index and searched on the design patent image database,build a content-based distributed design patent search system,and compared it through experiments the performance of system indexing in a single-machine environment and a cluster environment.On this basis,a distributed design patent search system with optimized process was proposed,and the two systems were compared horizontally.Experiments proved that the distributed design patent search system with optimized process can further improve the operating efficiency of the system.3)Optimized the design patent search system by adopting a multi-feature joint search strategy.This paper proposes a multi-feature joint retrieval scheme,which flexibly selects multiple features,calculates the distance between the image to be retrieved and each feature vector value of the image in the database,normalizes these distances and further processes them to obtain a retrieval result that integrates multiple features,this method avoids the limitations of single feature retrieval.Retrieval using single feature and multi-feature joint retrieval were carried out,and the experimental results were compared and analyzed.From the search results,it can be seen that the multi-feature joint search is more in line with the retrieval requirements.Experiments show that the distributed design patent retrieval system designed in this paper is efficient in processing large-scale data,and as the amount of appearance image data increases,the performance of the distributed retrieval system is more superior and can be flexibly expanded,meet the needs of users for fast retrieval.And the retrieval effect of multi-feature joint retrieval can better meet the rich and diverse retrieval needs.