The Research And Build Of Massive Pictures Retrieval Platform Based On Hadoop

With the constant development of information technology,how to achieve the storage and efficient retrieval of massive picture data,and retrieve valuable picture information that meets user requirements from these pictures,has become an urgent problem.According to the above problems,this paper studies the following aspects:(1)The thesis analyzes the development status and search results of big data storage technology,discusses the traditional Content Based Image Retrieval(CBIR)technology.On the basis of summarizing the characteristics and development of traditional picture storage and retrieval technology,this paper fully investigates the current mainstream retrieval algorithms.(2)The thesis implies bloom Filter algorithm to the problem of massive picture retrieval.The bloom Filter,as an irregular data structure of high space efficiency,its internal operation only requires simple "and" and "xor" operation,making the hardware relatively easy to achieve.At the same time,the matching efficiency is very high,so it is suitable to determine whether a file belongs to the set.On this basis,this paper proposes an improved mechanism,to reduce the misjudgment rate of the bloom Filter algorithm,and improve the retrieval efficiency.(3)The thesis uses Hadoop distributed file system HDFS and non-relational database HBase to achieve distributed storage and management of massive pictures.The Message-Digest Algorithm 5(MD5)value of the image is obtained by preprocessing the image and is as the RowKey value in the HBase database.Using bloom Filter to determine whether the image exists in the library.If the image is in the library,retrieve picture information in HBase according to the RowKey value.Build an integrated solution of massive pictures of big data based on HDFS-HBase-Bloom Filter,and the experimental verification is carried out.