Automatic Visual Gun Detection Carried by A Moving Person

Few works are reported in firearm detection, even though, it has an important application in both the field of video forensic analysis and surveillance. Security of different crowded areas such as around airport, marketplace, places of worships, shopping mall, ceremonies events such as marriage, parties etc. always have a requirement of automatic detection of moving person carrying illegal weapon. In the present scope of the work, we propose a gun detection technique based on template matching which is invariant to scale & rotation. Template matching is a simple and traditional method used for object recognition with a disadvantage of high time complexity. We propose a innovative yet simple way to reduce the time complexity by employing a background subtraction methodology. Background subtraction algorithm handles other challenges too, such as sudden change of illumination. Experimental results illustrate that the proposed method outperforms efficiently in gun detection in video sequences with lesser number of True negatives in comparison with state-of-the-art template matching algorithms.