Research On Safety Helmet Detection System Based On Improved YOLOv3

With the development of the times,some emerging technologies such as 5G technology,Internet of Things,cloud computing,big data,and artificial intelligence are booming,replacing some inefficient technologies in the past.The same is true for video surveillance technology.With the upgrading of technology,video surveillance is also moving towards intelligent direction,replacing the previously inefficient manual inspection methods and liberating productivity in a smarter and more automated way.In a complex work area,wearing a safety helmet is necessary to ensure workers' safe production.This article is based on the current artificial intelligence technology to provide an intelligent solution for surveilling workers wearing helmets,automatically monitoring workers in real time.And the system sets an alarm when not wearing to improve production safety and efficiency.Similar research has been done recently,but most of them are based on traditional detection methods.The actual application scenarios may not achieve the desired results.Generally,there are problems such as high time complexity,high environmental requirements,low robustness,and low real-time performance.This paper proposes a method based on deep convolutional neural network learning,real-time monitoring of camera video in the work area,and timely alarm when it finds that there is no wearing helmet,so that the background personnel can deal with it in time to prevent the occurrence of not safety production phenomenon.The method has many advantages such as real-time,correctness,strong robustness,etc.,which fully meets the actual production requirements.The detection system proposed in this paper is mainly divided into two modules,one is the detection module and the other is the tracking module.The detection module needs both rapidity and accuracy.On the basis of this analysis,the recent detection model is found.It is found that the YOLOv3 algorithm is in line with the system requirements,and it is decided to improve the YOLOv3 algorithm to make it more in line with the detection requirements of the helmet inspection on site.In order to prevent the occurrence of repeated alarms,a tracking module is designed.The DeepSORT algorithm is an improvement on the SORT algorithm.Under the premise of ensuring the speed,the situation of object occlusion is improved,and it meets the tracking application in the complex scene of the construction site.Therefore,in the algorithm designed in this paper,the tracking module is combined with the detection module to design a joint deep convolutional neural network algorithm model.After various tests such as pictures,videos,cameras,etc.,the system designed in this paper fully meets the industrial requirements in terms of real-time,accuracy,and robustness.It is very effective to intelligently detect whether workers wear helmets.