Dhaka University Journal of Applied Science & Engineering

Surveillance and security cameras are becoming much more common in city streets, shopping malls, private homes and many other places. In recent years terrorist attacks in shopping malls, schools and public places have increased. Such places, if equipped with a computer vision-based system which can effectively identify abnormal or violent human behavior, can help us in saving many human lives in due time. In this paper, we present such a system which can detect abnormal or violent human behavior in a video using machine learning. The system first identifies every human present in a video, extracts some vital data for that person and then, based on those data points, trains machine learning models to perform the classification. We have found that the system works reasonably well under certain conditions.

1. ref1