Research On Image Detection Technology Of Safety Protection Equipment For Electric Workers Based On Improved YOLOv5

With the continuous development of the times,the country’s power demand is increasing,and the security challenges to power workers are also increasing.Therefore,the research on image detection and recognition of power workers’ safety protection equipment is particularly important.Aiming at the low accuracy and poor robustness of the traditional detection algorithm of safety protection equipment in the power industry,and the problem that the original YOLOv5 algorithm may fail to detect or miss the detection of small targets,a new image detection algorithm of safety equipment for power workers based on improved YOLOv5 is proposed,and the improved algorithm model is transplanted to the Web terminal and Jetson Nano development board.The main research work of this paper is as follows:1)First of all,aiming at the problem of insufficient image data of workers’ safety helmets in the power industry,an image data set containing more than 18500 electric workers’ safety helmets was made by means of independent labeling.Secondly,in order to improve the safety of electric workers,a helmet image detection algorithm based on improved YOLOv5 is proposed.On the one hand,the feature fusion layer and multi-scale detection layer are optimized and improved,and a fusion scale layer for small target recognition is added;On the other hand,the K-means algorithm is used to re-cluster to generate a priori anchor frame that is more suitable for detecting small targets such as electric safety helmets,and CIOU Loss is also used as the loss function of boundary box regression.The experimental results show that the detection effect of the proposed algorithm for the helmet is better than the original algorithm.The average accuracy rate increased by 2.9% to95.0%;The recognition accuracy of safety helmet has also been improved by 2.4%,reaching 94.6%.2)First of all,aiming at the problem that the previous algorithm can only detect the helmet image,a set of 3104 images marked with protective equipment such as insulating gloves and helmets worn by electric workers during operation was made.On this basis,a multi-target detection algorithm model based on improved YOLOv5 is proposed,which can detect insulating gloves,safety helmets and electric workers in the image.On the one hand,it improved the backbone network,introduced the coordinate attention mechanism(CA)module,and improved the detection effect of small targets.On the other hand,the neck network is improved,and the bidirectional feature pyramid network(Bi FPN)structure is introduced to further improve the feature extraction ability.The experimental results show that the average accuracy is1.8% higher than the original algorithm,reaching 96.4%;The average accuracy increased by 0.4% to 93.3%.The algorithm model has achieved good results in the detection of insulating gloves,safety helmets and electric workers.3)First of all,the multi-target detection algorithm model of electric workers’ safety protection equipment based on improved YOLOv5 is transplanted to the Web side;In addition,in order to meet the needs of practical applications,through model compression and Tensor RT acceleration optimization,the deployment and transplantation of the model on the Jetson Nano mobile terminal has been successfully realized,and the detection speed can reach 24.3 frames/s.It solves the problem of time-consuming and labor-intensive in the traditional detection of electric power safety equipment,and can meet the requirements of real-time.The image recognition technology based on the improved YOLOv5 safety protection equipment for electric power workers studied can be an important part of the safety guarantee of the electric power industry,and is also of great significance to protect the personal safety of electric power workers.