Research On Early Warning Of Construction Workers’ Unsafe Behavior Based On Convolutional Neural Network

As my country’s construction scale continues to expand and investment continues to grow,the number of construction workers in the construction industry has gradually increased,and safety production accidents have also steadily increased.It is found from the investigation report of production safety accidents that accidents are generally caused by unsafe behaviors of people.Therefore,the effective identification of unsafe behaviors of construction workers can reduce the casualties and property losses caused by production safety accidents and maintain the health and sustainability of the construction industry.The key to development.Based on deep learning theory,this paper uses text mining and image recognition technology to identify unsafe behaviors of construction workers.It is divided into three parts: unsafe behavior feature word extraction,behavior recognition,and behavior early warning.First,starting from the production safety accidents in the construction industry in the past 5 years,the text mining technology is used to extract the feature words of the cause of the accident.The feature word structure is the location of the accident,the unsafe behavior that caused the accident,and the type of accident.Finally,the extracted feature words are classified and sorted to form a list of unsafe behaviors of construction workers.Then,according to the list of incomplete behaviors of construction personnel formed in the first step,the behaviors of construction personnel on the construction site are identified and judged whether the behaviors are safe.In the first stage,under the Tensor Flow and pytorch deep learning platform,use the python language to build the faster R-CNN and yolo network framework to identify the safety protection equipment of construction workers.In the second stage,the faster R-CNN network framework is used to extract and save the image information of the construction personnel,and the convolutional neural network is used to classify and predict the operation behavior of the construction personnel.In the third stage,personnel inspection in hazardous areas of construction sites.Use yolo network to identify and track personnel in dangerous areas.This paper uses target detection technology to identify three common types of unsafe behaviors of construction workers in three stages.After that,BIM technology is used to establish an early warning model of unsafe behavior of construction workers,which combines the detection of safety protection equipment,the detection of operation behavior of construction workers,and the detection of personnel in dangerous areas to build a BIM-based safety early warning model.Finally,the main conclusions and prospects of this article are discussed in three parts: extraction of unsafe behavior feature words,unsafe behavior identification and establishment of early warning models.The research in this article can effectively reduce the production safety accidents caused by human unsafe behaviors,and provide new management techniques and ideas for safety managers.