Detection and Tracking of firearms and cold weapons using deep learning networks

Since providing and increasing security in special and public places has always been the focus of everyone in various fields, many measures such as installing surveillance cameras and placing officers on the spot have been taken to provide and increase security in public places such as schools, universities, offices, and so on. Because these cameras detect suspicious activity by human operators, factors such as distraction, lack of awareness, fatigue, and a variety of other factors can affect the quality of surveillance; however, the firearms detection system detects and identifies defined weapons automatically. They can send special messages to military and security officials in a timely manner. In this paper, the Yolo version 5 algorithm is used to detect and track firearms (pistols) and cold weapons (knives), and the Deepsort algorithm is used for tracking. The article investigates the system's performance. Finally, after training with a special data set of firearms and cold steel, the proposed system performs firearms and cold steel detection operations with 99.50 percent accuracy. Furthermore, the proposed system has provided adequate accuracy in terms of scale, rotation, and obstruction.