Generated by socialized identification on web pages, pictures, videos and etc by the mass Internet users, tag is an unique meta data comprised of words on Internet. As a new data resource, it offers a new way to produce, manage and obtain Internet information. This paper will try to explore and utilize the characteristics of tag, start from hierarchical tag relation definition, integrate the knowledge of informatics and probability theory, and provide a easy and quick picture search method with relatively accurate data for users.