Research And Implementation Of Real-time Commercial Identification System

People have widely enjoyed the live TV service from the live TV provider such as CNTV in their daily life for recent years, and there are more and more kinds of live TV product that are covering all different channels played in provinces, cities and local places, the number of witch can be counted up to 140. The live TV show mainly composed by many different sorts of television shows, like news show, talk show, art show, entertainment show, documentary show, etc. it will be a wonderful achievement that we can use computers to catalog all those different shows efficiently and automatically in real time. Our system named shot video targeted real-time categorizing system, which can mark the commercial in live TV shows in real time, accomplished this wonderful task. After studied thoroughly on how to predict the category of a video clip and how to find out whether it’s commercial or not, we determine to choose the learning based method to classify the video clips and to utilize the practical application framework to obtain running efficiency. The whole system, taking live TV stream as input and displaying the output video stream with commercial clips labeled, can save the cost of labor works to label video files by hands and that means this system can take advantage of classification algorithm to make lots of profits for you.The content-based video classification is a most important subject in multi-media research. First of all, content-based video classification can be utilized in the management, storage and indexing of the video data, which is becoming more and more available in our daily life; secondly, by classifying video data, we can organize the video in the web efficiently and build the index tag of huge number of web video automatically to support retrieval of different kinds of video; last, we can take advantage of this technic to supervise the content of online videos, find prohibited video and prevent such illegal video to be spread out. In this paper, we made a big contribute to this kind of content-based video classification system and algorithm, and the result can be displayed in real-time.