Research on Automatic Classification Technology of Flash Animations based on Content Analysis

摘要

As a prevailing web media format, Flash animations are delivered and viewed by millions of Internet users every day. The search and classification technologies of Flash animations are important and necessary, but they are difficult due to the absence of understanding of the animation content and are not thoroughly addressed. By mining the file structure and content structure of Flash animations, an automatic analysis method based on the contents of Flash animations is explored, and some typical content features are extracted. These features include file size, graph number, image number, sound number, movie clips number, deformation number, button number, text number, script number, frame number and so on. Using these features, three common classification models, including classification and regression tree, neural network and support vector machine, are respectively selected to effectively classify Flash animations into 5 categories: game, cartoon, MTV, advertisement and teaching courseware. The experimental results show that the neural network model can make full use of various content feature information in Flash animations and has the best classification effect with 90. 26% of the average accuracy rate. Moreover, through conducting respectively experiments for the different categories of Flash animations, it is found that game is obviously different from others and they are the most distinguishable category of Flash animations. This research owns important reference value and practical significance in the content analysis, feature extraction, automatic annotation, intelligent search and classification management of Flash animations.