Adaptively feature learning for effective power defense

Abstract

Active defense technology is very important in intelligent systems and video surveillance. In some important fields, active defense system can effectively find intruders. Many intelligent video surveillance systems were proposed in recent years. They achieved good performance to some extent. Since power station is an important field, it is important to develop an intelligent video surveillance. Considering that detecting the whole surveillance video is time consumption and computation. So in this paper, we propose an active defense system to find intruders automatically. First, a key frame selection algorithm based on adaptive features is presented to select key frames. These key frames can cover the main content of videos and detecting these key frames can also reduce time consumption and computation. Then, a probabilistic model is proposed to learn the training data distribution. Finally, our system can achieve active defense based on probabilistic model. Experimental results show that our active defense system can achieve finding intruders effectively.