Design And Implementation Of The Violent-terrorist Video Recognition System Based On Logo Markers Detection

With the spread of violent-terrorist video in the network, how to quickly and accurately detect those terrorism related violence video and establishment of a safe and healthy network environment is an important issue to the whole community. In order to build a multidimensional terrorism related violence video detection framework, the main research work is based on multiple-cue fusion of video semantic event detection model and based on this model for the network environment of multi dimension violent video detection framework.Violent terrorist video’s Logo is used to identify an important symbol of the video content, It plays a very important role in the detection and identification of violent video. Along with the request of the national information technology is becoming more and more high, the detection and recognition of Logo in the video has more and more application prospects in the information security as well as maintain social stability and so on.This paper at first further study and analysis of the current status of video Logo sign detection and recognition, analyzes and summarizes the specific problems affecting Logo detection and recognition in the practical application exist. Then propose a method based on template matching detection to recognition terrorism related violence video’s logo, analyze the different classification of template matching method and different matching methods are suitable for the different application scenarios. Finally, use the gray template matching algorithm according to the video logo features to realize the detection of dynamic video logo.Then study and analyze the classical algorithm MAD、SAD、SSD、NCC and SSDA according to the gray template matching, on base of these, propose a fast template matching algorithm in view of the defect that the time complexity is relatively high, which is suitable for the detection of violent terrorist video Logo. Through the gray value coded to rough match, and then precise match using phase correlation method, so that the speed of template matching is improved.In addition to the search strategy, there usually have a general shape and size stability for single video logo, through increase of the key frame step size, it also greatly improves the speed of video Logo detection, and the experimental results are also very obvious.In the design process of the whole video logo detection system, every step of the reasoning are closely linked to the video logo background, and the selection of correlation algorithm is also in accordance with the practical application. The experimental results show that the Logo detection system from the special sign has a great improvement in the recognition accuracy rate and false detection rate of the terrorist related violence video detection system. and the contribution to the final result is also very obvious.