The Magazine Cover Recognition System Based On Mobile Terminal

With the development of the Mobile Internet and the massive popularity of mobile terminals, the traditional cardboard magazine has become unable to meet the demand of the general public. The image not only is an important information-carrying form, but also is an significant factor to attract users and readers to interact with media.Users can photograph a magazine cover with a mobile terminal, then use the image processing technology to help users to achieve the information query, dynamic tracking and analysis of the image transmission. The key technique during the process is the image recognition, therefore, how to accurately and effectively identify the target image from a large number of images with the mobile terminal has become a significant research topic.Due to the wide range of existing magazines with complex names and LOGOs which are difficult to identify,the inconvenience of text input mode, and the low efficiency of traditional matching algorithm, this dissertation studies in depth the related technology of image recognition, presents a matching algorithm based on the magazine covers’own characteristics:firstly, locating the rectangle of the cover, secondly, recognizing the magazine categories by using the masthead LOGO’s SURF feature, finally, recognizing the specific magazine cover according to the SURF feature of the cover’s unique background image. In addition, the nearest neighbor method and the RANSAC algorithm are used to eliminate the mismatching points. Hence, this algorithm could greatly improve the accuracy of the image recognition for the cover of the magazine.According to the modified algorithm, this dissertation also actualizes the magazine cover recognition system using the C/S mode based on Android mobile terminals.In other words, the mobile terminal captures the magazine cover image and transforms it to the server,which can uses the improved image recognition algorithm to identify the image, then feedback to the mobile terminal, thus to get the relevant information of the magazine. With a large number of real cover images under the conditions of illumination, rotation and interfering backgrounds, the dissertation compares the accuracy as well as the efficiency of forementioned algorithm to SIFT and SURF. It can be proved that the verification of this system has better recognition accuracy and robustness. Thus, we may draw the conclusion that the magazine cover recognition system the magazine recognition is well applicable for mobile terminals.