Detection of fence climbing using activity recognition by Support Vector Machine classifier

Detection of person climbing the fence is a vital event, taking borders and restricted zones securities into the consideration. In this paper, a model is presented which detects the human crossing a fence. Entire event of climbing a fence involve activities like walking, climbing up and climbing down. Moving person is detected by background subtraction algorithm. Centroid of the blob and centroid variations along the frames are considered as features. Support Vector Machine classifier is used for detection of walking, climbing up and climbing down activities. The experiments show the best results in detection of human crossing a fence compared to the existing state of art methods.

I. Introduction

In present scenario, video surveillance has an important role in security purpose at various locations like airport, railway stations, shopping malls and restricted zones. The requirement to monitor more persons, places, and things along with a desire to extract out more and more useful information from video data sets is motivating latest demands for capabilities, scalability and capacity. There are immediate requirements for fully automated video surveillance systems in law enforcement, commercial and defence applications. These systems will help in assisting the human operators.