Research On Intelligent Safety Inspection System For Power Workers Based On Image Recognition

In electric power construction,safety is the most important and basic requirement for workers,and electric power construction activities are one of the typical high-risk production activities.In recent years,with the development of cities,the scale of the power system has continued to expand.Safety accidents have occurred frequently in the process of power operation and construction and maintenance,and personal accidents have been repeatedly banned.Accidents often occur because of the negligence of workers,assembly tools and non-standard behavior.In recent years,artificial intelligence technology has been gradually applied in various fields of life,providing new solutions for the realization of intelligent management of electric power construction sites.However,the function of many construction sites currently is relatively single,only wear parts recognition or face recognition,and there is no further processing of the identified data to extract the behavior characteristics of workers.At the same time,with the expansion of the scale of the construction site,unified management of the construction site has become a demand.Managers need to use limited resources to focus on inspections of high-risk construction sites,thereby improving the management efficiency of multi-sites.At present,there is no system that processes the data collected in real time on the construction site to extract site features and evaluate the risk coefficient of the site.In order to meet the above requirements,this dissertation designs an intelligent security detection system for power workers based on image recognition.The system collects the camera data of the construction site,and transmits the video stream to the safety helmet detection module and face recognition module,so as to realize the safety helmet recognition detection and attendance function of workers.The safety helmet detection module adopts the target detection algorithm based on deep learning,and the face recognition algorithm also adopts the method of feature extraction and feature comparison based on deep learning.In the target detection network of safety helmet detection module,an improved YOLOv3 model based on CBAM attention mechanism is proposed to judge whether electric power workers wear safety helmet,which improves the recognition accuracy of safety helmet;In the face recognition module,the face recognition architecture based on MT-CNN & Facenet is adopted to realize the correct recognition of workers identity.At the same time,a face recognition network based method to count the working hours of workers is proposed to prevent the workers from fatigue.When the construction sites are united,the data collected by the system and the environmental data of the construction site are further processed.The multiple linear regression model is used to evaluate the risk of the construction site,and the risk coefficient is sorted from high to low,so as to provide reference for the local decision makers to supervise and investigate the potential safety hazards of the construction site.