Intrusion detection of specific area based on video

Abstract

In order to solve the problems of object detection and object tracking under complex scenes in video, this paper proposes a way to improve Gaussian Mixture Model algorithm based on the traditional Gaussian Mixture Model. When the model is updated, according to the characteristics of continuous video frame, the background model is divided into static regions and dynamic regions, and the background is updated in different strategies. Then, this paper presents an algorithm for the intrusion detection. Intrusion is judged by whether the centroid of the target is in the specific area. If the centroid is located outside the area, it shows that the target does not invade the specific area, otherwise the target invades the specific area. If so, the system triggers alarm and label information appear on the video frames. Experiments show that this algorithm can realize the intrusion detection of specific area.