Novel Deep Neural Network for Suspicious Activity Detection and Classification

Implementing preventive measures against gunshots and terrorist attacks in public areas with heavy foot traffic is crucial. Although security cameras are already commonplace, real-time reaction and 24/7 monitoring needed for automated detection techniques. As opposed to a weapon alone, a weapon in the hands of a human is thought to pose a larger threat. This paper has introduced a novel algorithm for multiple gun detection with an innovative deep learning neural network (DNN) model. The parameter for human suspicious activity in this paper is a person with a weapon(gun) and a person wearing a helmet with a weapon(gun). The temporal complexity of the Proposed System Architecture (PSA) on various hardware platforms is also explored. The algorithm gives 99% accuracy as compared with other existing methods.