| With the rapid development of internet industry,new technologies and applic ations occur continuously.The internet information which regards pictures and vi deos as its carrier promotes the dissemination and communication of information to a large extent,internet information is gradually taking place of traditional me dia and it becomes the media form with the highest usage rate.When bringing people with convenience and erasing the information gaps,in ternet information also produces some unhealthy information.Recent years,gover nment authorities have made a lot of work on cleaning up the internet space,es pecially in the aspects such as market access,content requirements,daily adminis tration and industrial self-regulation etc.,these all-round methods have been taken by government authorities to strengthen the management of internet audio-visual programs,they clean up the harmful and unhealthy programs and the piracy pro grams resolutely and spare no effort to clean up the illegal networks and illegal internet TV programs[1].However,some problems still exist,for instance,some n etworks and content service providers seek for economic interests in a none-side d way,they don't check the contents strictly,and they disseminate vulgar content s,illegal contents and fake contents.For example,some content service providers follow the public to transfer internet rumors so as to increase the dissemination of internet rumors,and they bring potential or apparently negative influence to the society.The aim of the study is to design and realize a set of big data environment oriented video image recognition system,taking NoSQL database design ideal as its foundation.According to the stock current status of internet images and the high handling capacity of business system on videos and images identification,as well as the essential demand of high stability and low cost deployment,the sys tem realizes a video image identification system with low coupling of blocks,hi gh robustness of functions and abundance deployment of supportable distribution.Meanwhile,the system takes the difficulty of current internet data obtainment in to consideration,since most of video resource figures can't be dominantly downl oaded and it brings big trouble to the business system,so the system merges dat a collection function into the system design,it adopts web crawler block to colle ct internet data on its own and promotes the competitive ability of the business of the system.In addition,on the aspect of identification cost,the system adopts the way of self-protection on black and white list library continuously to avoid the waste of repeated calculation resources of data by studying on a large num ber of internet figures.The cost of system deployment is decreased in a large sc ale.The main work includes:deeply study on the system design and realization under the background of big data,construct the entire idea of constructive syste m design based on NoSQL design ideal,choose appropriate NoSQL database as the base of system realization;analyze the distributed file system,according to t he current status that most of the internet figures focus on little files,choose FA STDFS distributed file system to store efficient figures;adopt docker application lawyer virtual engine to fix the function programs and allocation of each block,realize the high portability of system and liberalization deployment,finally use s ome program languages such as JAVA,PYTHON etc.to accomplish the develop ment of image identification system.The characteristics of the system include:based on the ideal of NoSQL,the system designs of a set of rapid identification system of images that meets to t he demand of NoSQL database quality,it has the features such as horizontal exp ansion system quality,supporting redundant deployment of multiple dots,high po rtability and reliability,complete business identification procedures and relatively low deployment cost etc.,it satisfies the business demand of image identification in the environment of big data. |
PDF Full Text Request