Design And Implementation Of Illegal Action Recognition System In The Scene Of Electric Power Infrastructure

Safe production is an important foundation,premise and guarantee for the daily work of electric power enterprises.Infrastructure construction is a high-risk field of safety production.High-risk operation scenarios such as close-to-electricity operations and cross-over construction on the infrastructure construction site have relatively large safety risks.It is easy to produce blind spots in the safety management and control of the operation site.Therefore,strictly formulating and effectively implementing safety risk management and control measures at infrastructure construction sites is the key to ensuring safe production.By combining the national "double carbon" goal,the construction of new power system of State Grid Corporation and the strategic thinking of "digital new infrastructure",this paper introduces artificial intelligence and digital technology to the process of infrastructure construction,deeply integrates "artificial intelligence" and "construction site safety risk management and control",and strengthens the safety risk management and control of the construction site.This paper systematically analyzes the problems existing in the safety risk management and control of the infrastructure construction site,and proposes specific measures to improve the safety risk management and control.A technical support system has been established to ensure the effective implementation of security risk management and control.The electric power construction site violation identification system has been developed.In view of the rich types of infrastructure site behavior data,the difficulty of monitoring violations,and the low efficiency of safety risk management and control,the design and research is carry out in three aspects,i.e.,the construction of infrastructure site behavior data sets,intelligent analysis methods for violations,and infrastructure site violation identification systems.Firstly,this paper analyzes the static and dynamic characteristics of the on-site operation behavior data of electric power infrastructure,combines with the behavior recognition public data set,designs and constructs the infrastructure on-site behavior data set,and studies the data preprocessing technology of illegal behavior in different scenarios.Secondly,this paper proposes a deep learning-based identification method for on-site operation behavior of electric power infrastructure.By introducing the time-shifting idea and attention mechanism module on the backbone network,this paper constructs a backbone network structure including space-time representation and space-time movement information on the channel and space,realizes the space and time feature extraction,introduces of the attention module to enhance the ability to express the key details information of the network,and improves the network’s ability to extract important time-series action features.Thirdly,this paper analyzes the application scenarios and requirements of the on-site violation identification system for power infrastructure,and designs the network deployment architecture,system architecture and database of the on-site violation identification system,has realized the functions of system label and data management,data download and violation behavior analysis,has completes the function and model testing for this system.The system has been applied to the electric power infrastructure site,and has realized illegal behaviors identification effectively.