Research And Implenmentation Of Web Advertisement Image Filtering Technology

Since the 90s of last century Internet access to China, Internet penetration rate of Chian has reached 51.2%, the size of Internet users has reached 710 million, more and more people release or obtain information form the Internet. Such a large group, contains a huge business opportunity. Which result in Web pages are filled with more and more ads. This phenomenon caused a serious impact on what people get useful information form web. Since entering the Web2.0 era,because of the image has better visual effects and can contain richer content, images are increasingly used to spread advertising messages that seriously affected the people's work efficiency. At present, there are a lot of filtering research on the image of advertising,but most of the research is based on the specific content of the image classification and recognition, although the accuracy is higher, but the image recognition is difficult, the algorithm is complex. Consider this situation, this paper studies how to effectively carry out image filtering.The work as follows:1. This paper summarizes the characteristics of advertising images,analyzes the advantages and disadvantages of the current feature selection about image and combines the characteristics of Web ads with personalized interest-oriented. The feature extraction is carried out from four aspects, such as interest, text, links and attributes. I propose and implement an image filtering model based on SVM.2. I study the HTML text of DOM attribute, combining the characteristics of advertising images and current user interest based advertising recommendation. I study the advertising image filtering technique based on DOM attributes and avoid the recognition of image content, then I propose a model based on the four aspects of interest,text, links and attributes of a total of 11 features. I verify the validity of the model form accuracy, precision, recall and F1 measure.3. In the extraction of text features, I study and comparative the commonly used keyword matching algorithm. Considering the matching content in this paper is easy, I choose the forward maximum matching algorithm.4. In this paper, I learn the HTTP transparent proxy and Internet Content Adaptation Protocol. I propose an advertising image filtering system based on the DOM attribute on Squid-ICAP, introduce the design of the system and the design and implementation of the key function modules. Finally, I verify the validity of the system.