Research And System Application Of Industrial Factory Safety Helmet Detection Model Based On Deep Learning

In recent years,with the in-depth and rapid development of visual target detection technology and machine learning,deep learning artificial intelligence as the technology-based target detection technology field has made many major technical discoveries and is gradually developing into the modern industrial computer vision A mainstream detection algorithm.However,with the development of deep learning artificial intelligence and computer deep learning detection technology,video security monitoring is also moving towards more intelligent and more in line with the technical direction of automation.In the automated factory operation area,wearing a helmet inspection is a necessary condition to ensure the safe work and production of construction workers in the factory area.Based on the above analysis,this article can propose a detection technology for video helmets based on artificial intelligence deep learning,which can carry out real-time supervision of the work of construction technicians wearing video helmets in the industrial plant area and pass through the industrial plant area.The real-time alarm of the camera of the person who did not wear the helmet in time,thus effectively protecting the production personal safety of the construction workers in the factory area.The main tasks and contents of this article are as follows:1.This article first studies the basic theory of deep learning,and then analyzes and studies Yolov3's network architecture,detection process,and loss function.2.Apply the target detection algorithm to the target detection task of the helmet in the factory area.In view of the balance between accuracy and speed,Yolov3 is selected as the basic network in this paper,and 60 channels of surveillance video are collected on the spot as the training data of the network.3.In order to further improve the accuracy of the network,this article replaces the original IOU of the network with DIOU,and then replaces the traditional NMS with Softer NMS.In order to improve the generalization ability of the network,this article has carried on the data enhancement to the data.4.In order to greatly reduce the calculation and deployment management costs of the forward network data propagation model,this article uses TensorRT to lighten the network model to accelerate the processing operation,which greatly reduces the network resource consumption of the calculation of the network forward data propagation model And occupancy,which greatly reduces the cost of network model deployment management.