Research On Safety Monitoring Of Personnel Behavior In Construction Site Based On Computer Vision

The construction industry is a high-risk industry,and construction safety accidents occur from time to time,which has a lot to do with the characteristics of the construction site.In the complex construction environment,there are multiple types of work at the same time,the flow of personnel is large,and the simultaneous operation of different machines increases the safety risk.Analysis of statistical data shows that human error is one of the main factors that cause production safety accidents in housing and municipal engineering in my country.A large number of construction safety accidents have caused irreversible loss of life and property and bad social impact.At present,emerging technologies are developing rapidly.Based on the rapidly developing computer technology,sensing technology,signal processing technology,artificial intelligence and other technologies,using the computer as the operating platform,the automatic,real-time and fast construction site monitoring system has entered people’s field of vision and been extensively studied.This allows us to process monitoring data quickly and accurately with the help of new technologies,conduct long-term online monitoring of the safety status of complex construction sites and buildings during construction and operation,timely discover various hidden dangers and provide quick feedback to assist in management Personnel to supervise and ensure the safety of the construction site.This paper mainly studies the following aspects:(1)A method for target recognition in complex backgrounds based on generative adversarial neural networks and convolutional neural networks is proposed.The two neural networks are used in combination,and the generative adversarial neural networks are used to preprocess the data.Complex backgrounds are optimized,simulated data are generated,and high-quality datasets are obtained.Then use the convolutional neural network to detect the specific behavior in the concrete construction process.Through comparative experiments,this method can effectively improve the average accuracy of target detection.(2)A method of human action behavior recognition based on human skeleton joint point pose estimation and convolutional neural network is proposed,which can improve the recognition accuracy of small objects with typical behaviors.This part is divided into two parts.The first part introduces Open Pose,a classical algorithm for estimating the pose of human skeleton joints,and uses this algorithm to estimate worker behaviors and screen out dangerous behaviors with typical movements.The second part introduces the convolutional neural network Faster R-CNN algorithm,and uses this algorithm to further detect the filtered data and identify the dangerous behavior of people.(3)Finally,based on the above research content,a method of building a real-time monitoring platform based on construction site monitoring video and machine vision related technologies is proposed.On-site management,improve work efficiency.A data storage optimization method based on convolutional neural network is also proposed,which improves the utilization of storage space.This paper focuses on the unsafe behavior of construction site personnel,explores the monitoring methods for unsafe behavior of construction site personnel,builds a monitoring platform for intelligent monitoring,provides necessary assistance for construction management personnel,and conducts digital management of the construction site from all aspects and angles.The intelligent management of construction sites is of great significance.