|
In the past decades, traditional relation database manage systems play a very important role in commercial data management. But, with the development of computer application, especially, the evolution of Web information technology, the data has presented two prosperities, magnanimity and sharing and its characteristics are diversiform, thus, less traditional structural data, but more un-structural data, such as image, media, video, document, emerge in people’s lives. However, traditional database cannot handle these new prosperities of data, so people propose the definition of dataspace.Dataspace can solve the problem of low efficiency in the query of personal data management, such as failing in dealing with un-structural data, keywords query and association query. Accordingly, in this paper, we research the technologies of content based image index in dataspace. Based on M2-tree, we propose a more efficient multi-dimensional index, M2+-tree. It put forward a novel definition of key vector. After the partition of the whole space based on Euclidean distance, M2+-tree partitions the whole space again, by the rule of key vector. While the range query and K-NN query, it first filters unnecessary data, according to triangle inequality, and then filters more subtrees that do not contain any query results, by key vector filtering, which improve the query efficiency. What’s more, we also propose bulk loading algorithm of M2+-tree, which significantly enhance the efficiency of it.Finally, using M2+-tree, we implement content based image index, in a dataspce system, named Mydataspace. This system provides extracting image characteristics, establishing M2+-tree index and query interface. The user can get a similar image comparing with the image he or she imput.
|