Research On Power Construction Security Monitoring Technology Based On Image Analysis

Power construction is a necessary means for the development and progress of power grid companies.Therefore,safety monitoring of power construction has also become an important research topic for power grid companies.Traditional construction scene safety monitoring uses manual methods to monitor,which not only wastes a lot of human and material resources,but also cannot guarantee real-time monitoring.This paper designs a power construction security monitoring system based on image analysis.It mainly studies the problem of target detection on the power construction site.The main problem to be solved is whether workers are wearing safety protective gear correctly in the power construction scene.Intelligent monitoring based on images is adopted.It can realize the whole process monitoring of workers ’construction process and ensure the safety of worker.This topic mainly studies the target detection technology in the power construction security monitoring system.First,use the images collected by the on-site monitoring system to make the dataset.During the production process,select valid data for sample labeling to obtain a dataset suitable for training and testing of power construction scenes.Second,for the complex working conditions of the power construction site,In the case of many small targets,a multi-branch network architecture based on the Darknet-53 basic network is designed.Gio U is integrated in the NMS,and the network structure and detection principle are introduced.Finally,experiments are performed on the designed network.Using self-made data set based on electric power construction scene for training,a stable and reliable target detection model is obtained,and the experiment of target detection in electric power construction scene is realized.In this paper,the designed multi-branch network structure detection model is compared with the YOLOv3 algorithm.Experimental results show that the model can effectively locate and identify targets in power construction scenarios,with an average accuracy rate of more than 95%.The detection speed on the RTX2080 Ti can reach 106.5ms per frame,which can be applied to the power construction site to achieve intelligent monitoring of the construction site.