A Lamdmark Retrieve System Based On TASPT Algprithm

With the computer technology and Internet technology, and recent digital cameras, mobile phones cameras popularity of electronic products, coupled with the rapid development of the tourism industry, a large number of Internet users every day to see his shot or meaningful Pictures published or shared on the Internet, on the subject of image resources and day to surge. According to statistics, Facebook last October number of images has been more than ten billion.The face of such abundant resources, a growing number of computer vision researchers will study the scope of focus to the Internet-based image retrieval. According to the different retrieval methods, image retrieval can be divided into two types, namely, text-based image retrieval and content-based image retrieval. Text-based image retrieval system using artificial images the content on the label text messages, text messages using DBMS storage and retrieval technologies text query. Content-based image retrieval uses visual content of the image itself to be indexed. Existing image content-based image retrieval mainly used physical characteristics, such as:color, shape and texture.This careful study of theory and method of image retrieval, image retrieval focuses on the implementation process and the main algorithm. Learning through research of existing algorithms, is proposed based on physical characteristics of pictures and text labeling combined TASPT (Bind of Text Annotation and Spatial Pyramid Transform) algorithm. The algorithm set collection, training and computing image similarity, and several landmark steps in the use of text annotation and indexing of ways to help the image acquisition and retrieval. This landmark images in the calculation of the similarity transformation using the method of spatial stratification combined with a text correction in the speed and accuracy has been improved to some extent. Establishment of a landmark article TASPT image retrieval system, according to the similarity of the target pictures to sort and return the result chart. In this paper several experiments on this system and the experimental data and results of a detailed system performance analysis, and ultimately the corresponding conclusions.