Research On Automatic Recognization Of Specific Video Content

In recent ten years, the rapid growth of network greatly enrich people’s information amount. In the mean time of bringing convenience to people, it adds much complexity to identification work. Therefore, automatic identification system based on computer technology comes with the tide of fashion, it can effectively overcome the shortcomings of low speed, high error rate, enormous human resources consumption of artificial identification. The automatic identification method of certain video content this article discusses is mainly based on content feature of image, including two major parts:auto identification of station symbol and scratch defect frame.The system mainly provide help to copyright administration. Nowadays network has been one of the master stream media, and it’s connection with traditional TV broadcast channel is becoming more and more closed. The easy spreading way has also made abundant data redundancy and inconvenient to copyright administration. The system will achieve real-time auto identification through real-time division, feature extraction and comparison of video information which helps administrating copyright much more easily. Auto identification system of station symbol includes two parts:template extraction module and objective matching module. At the first step, extract all the target video modules and save them into database ahead of time, the improved background minus division algorithm: spatio-temporal slice algorithm adopted in this article, extracts binary template of station symbol which is clear and semi-transparent. After that, intercept real-time video stream from network or broadcast television system, with the coordination of" dynamic adaptive window, relative topology matrix extraction and method that improve the algorithm accuracy (logical and operation and siding window),the circumstance of dynamic resolution is effectively adapted, and it will enormously reduce data redundancy of template base. Besides, the lower time complexity ensures real-time operation of system.Identification system fixes those scratch defects made by physical cause through digital method, it can reserve age-old literature and history video material and etc much better. It is divided into two parts equally:detect and repair of defective frame. It identifies the character of different operations via analyzing different edges, which adopts Canny operation to do edge detecting work of image. Afterwards, boundary tracking method is used to identify straight line in binary edge image. Finally, the fake scratch is removed through gray morphological operations, the position and length of scratch are precisely located. In scratch repairing module, it’s mainly realized by bilinear interpolation.