Detection of Suspicious Activities of Human from Surveillance Videos

Abstract

Video surveillance has been used from a long time to provide security in many sensitive places, so with this great progress in various aspects of life the traditional surveillance operations are facing many problems because of the large amounts of information that must be handled manually in a limited time also the possibility of information loss which can contain important things such as suspicious behaviors. So recently, a large amount of research has been conducted on video surveillance. This paper will present a system to support the smart surveillance for detecting abnormal behaviors that represent security risk. The proposed algorithms are intended to detect two cases of human activities namely, walking and running. No restrictions were imposed on the number of people in the scene, and the direction of the motions. However, we restrict the videos to indoor color videos, where the video are captured by one stationary camera. The moving objects which correspond to people in the scene are detected by background subtraction algorithm . We consider the displacement rate of the centroids of the segmented foreground areas and the rate of change in the size of the segmented areas as the two main features for activity classification. Briefly, in this study, a sequential of procedures were applied to detect moving objects (suspicious activity) in video, these procedures consisted in dividing the video into frames and separating the background of the video from the objects inside it, as well as removal the noise from the images by using morphological operations thus the mathematical operations are performed to which to determined which images contains suspicious activity. The proposed algorithms determine the activity type with a high accuracy rate.