The Design On Re-ranking Video Retrieval Cloud Platform Based On Hybrid Method

With the rapid popularization of Internet and the rapid development of the Internet terminal equipment,the mainstream of video data also changes from the traditional text form into a rich multimedia form.The number also appeared a massive increase in the trend.The traditional video retrieval platform is also faced with the problem of semantic gap and lack of video content.They can't effectively understand the semantic information of query words,and also ignore the rich video content information.The massive and heterogeneous video data,efficient video management capability and massive storage capacity are the urgent challenges that facing the industry.In this paper,aimed to solve the problem of semantic gap and the problem of the missing content in the traditional video retrieval platform,a design on re-ranking video retrieval cloud platform,which is based on hybrid method is proposed.The platform is based on the semantic filling technology,the visual word bag technology and the distributed file system storage technology.The goal of this platform is to improve the user's satisfaction,and provide better technical support for video information management.The main work is shown as follows:(1)The overall architecture of the re-ranking video retrieval cloud platform which is based on hybrid method.Combined with a variety of video retrieval optimization method,this paper proposes a reranking video retrieval cloud platform,the architecture of the platform is divided into four modules: distributed video management module,bag of visual words module,visual and text feature correlation calculation module,calibration algorithm module.Respectively to realize video distributed management,video image processing,video retrieval and calibration functions.They mainly include the distributed management function,the image processing function,the retrieval calibration function.(2)The research on the video retrieval method based on text semantic filling technology.To solve the problem of semantic gap in video retrieval process,this study involves the technology of semantic filling based on open data,the work of this stage mainly includes filling the video label with semantic information,querying similar semantic concept,showing the semantic web graph of the video labels.On the basis of the traditional text retrieval,it is not only a simple word matching,but also the key to fill the words with semantic information.(3)The research on the video retrieval calibration method based on the bag of visual words model.To solve the problem of the video content missing,we proposed a re-ranking video retrieval method based on the method of bag of visual word,which combines the technology of extracting the visual characteristics of the video data,TF-IDF technology,open data technology.To provide users with the optimized video retrieval calibration results.First,it is need to use the key frame clustering algorithm to get the weight vectors of the key frame set;then using the Speed Up Robust Features of the key frame image to present video content features,this is the solution of the problem of video retrieval content missing.(4)The implementation and verification of the re-ranking video retrieval cloud platform based on hybrid method.This paper mainly uses our method to implement the platform.aiming at the problems of massive storage of video data,we use the cloud storage technology which is based on distributed file system to solve the problems of fault tolerance,big data set,linear expansion,and data backup.Finally,after a rigorous evaluation of functionality and performance,the results showed that the re-ranking video retrieval cloud platform which is based on hybrid method has a very good innovative value in the field of video management.