Violent Action Recognition Using Drone Surveillance

Abstract

quotIn today s society, there are an increasing number of incidences of-reported and unreported newlineviolent activity. In response to this growing danger, monitoring system fills the void of unreported behaviours that lead to violence. Drone and closed-circuit television (CCTV) surveillance footage can record a wide range of real-world oddities. The advancement of automated newlinevideo surveillance research is influenced by violence detection and facial identification of the newlinepeople participating in the violence. Drones and computerised video surveillance are becoming newlineincreasingly popular as a result of rising threats in society and a lack of personnel to monitor newlinethem. Violence detection may be used to filter surveillance footage and identify or note the newlineindividual generating the anomaly. Different strategies for recognising violent actions and identification of violent individuals were newlineexamined in this study. The initial work examined the effectiveness of a violent action recognition model that distinguishes between two, eight, and fourteen different violent actions. Further, newlinethis work was extended to utilise the transfer learning-based network architecture for violent interaction as well as aggressive individual recognition in drone surveillance video. A graph newlineconvolutional network was also examined to classify violent actions. The indicated systems newlinerun-time effectiveness was also examined for real-time performance.quot newline newline