Research On Safety Helmet Wearing Detection

With the development of society and the progress of science and technology,researchers have carried out a lot of in-depth research on computer vision,image processing and pattern recognition technology,and applied them to all aspects of modern social life.As an important safety tool,safety helmet can protect the head safety of construction workers.It is required to be worn correctly in various construction sites.In recent years,a large number of safety accidents have occurred in various construction sites,most of which are caused by some violations by construction workers,and incorrect wearing of safety helmet is one of the most important ones.Therefore,it is necessary to design an advanced method with real-time,accuracy and predictability to monitor the safety helmet wear of construction workers in real-time,in order to effectively curb this phenomenon.In this paper,the use of safety helmet in various construction sites is analyzed,and the wearing detection methods of safety helmet for construction workers are studied in depth.In this paper,the following difficulties exist:(1)The construction site environment is complex and there are many people.(2)It is difficult to obtain data of construction site.(3)Safety hats have different styles and colors,and are susceptible to environmental impact.In order to solve these difficulties,firstly,the advantages and disadvantages of various moving target detection methods are compared,and then the key point detection method is selected for human target detection.This method can simultaneously detect multiple human key points,and has good robustness and real-time performance.It can well deal with the problem of numerous and real-time monitoring on the construction site.Compared with various target detection algorithms,YOLO algorithm is selected to detect the position of human body frame and helmet,and get the position of human body frame and helmet.This algorithm regards detection as a regression problem,and gets the target position and category from the whole picture information.The detection accuracy is higher and the detection speed is faster.Then,the center of the head region is determined by the key points of human head,and the human is used to get the center of the head region.The shorter side of the body frame determines the length of the frame of the head area and obtains the human head area.Finally,the relationship between the human head area and the position of the safety helmet is judged to determine whether the safety helmet is worn correctly.In this paper,in order to study and use,through the simulation of construction site situation,a data set is collected and produced.The experimental results show that the safety helmet wearing detection algorithm studied in this paper basically meets the detection needs of construction site,lays a foundation for the application of safety helmet wearing detection algorithm in construction site after that,and also provides a reference for the follow-up research of safety helmet wearing detection algorithm.