Character Image Retrieval Based On Intelligent Terminal

With the development of the mobile internet, a growing number of new medias emerge and erode the market of the traditional media at a high speed. Some traditional media treat the prosperity of mobile internet as challenges and opportunity wisely. New media has the characteristics of simple, interactive, instant and flattening, while the traditional media own the advantage in creating articles. As to traditional media, it is an inevitable trend to combine the technology of mobile internet and the advantage of itself on the way of digitization.According to the issues mentioned above, this topic brings up a plan to help the traditional media to make use of mobile technology. In this plan, we develop a system that composed of a serious of technologies in the field of pattern recognition. Such as the image recognition, character recognition, information retrieval and public opinions analysis. We separate the system into three parts. Firstly, it’s the document recognition. In this part we use the algorithm based on local image patches to obtain the basic information of the media. Secondly, we provide an interface for the users in region selecting on the touch screen of terminal. And this part will translate the text image into sentences, so the computer will understand what the user want to know. Finally, the system will slice the sentence into key words. Then make retrieval and sentiment classify in a resource library with the words.On the analysis of theoretical in related field, this paper brings up a new method, combine the stroke interval density and the directional element feature to extract features from the character image. As well this paper designs a module in sentiment classification.In the forth section of this paper, test cases designed, result analyzed and the running systemshown in very detail. The component parts of the system all achieve the goal set before test.