A Staff Violation Detection Algorithm Based on Lightweight YOLOv3
With industrial Internet security receiving more and more attention,how to ensure information security manage-ment through technical means has become a question for unit managers to consider.In this paper,a lightweight convolutional net-work-based algorithm for staff violation detection is proposed for the security needs of similar scenarios such as duty rooms,central control rooms and management rooms.The algorithm is able to monitor the behavioural status of staff in designated areas in real time,avoiding malicious incidents such as production accidents in equipment and facilities and leakage of equipment information that may occur when staff are away from work or sleeping for long periods of time.The violation detection algorithm consists of two parts,which are the human detection and the behaviour recognition algorithm.Firstly,the human detection frame is obtained through a lightweight human detection network.The human detection frame is then identified using a target tracking algorithm and a behaviour recognition algorithm to determine if there is a breach of the rules by the staff member.Experimental data shows that the algorithm significantly reduces network weights as well as computational effort,with detection speeds of up to 13 ms on edge devices and violation detection accuracies of up to 96.6%on real-world scenario datasets.
国家科技期刊平台
NSTL
万方数据
