Exploration Of Safety Problem Discovery And Deep Learning In Construction Site Based On BIM And UAV Technology

For a long time,safety accidents in the field of construction occur continuously,which consume social financial resources,material resources and even take people’s lives and property.It has become an urgent problem for safety managers in the industry.Researchers also continue to innovate and explore,hoping to find a new safety management method applied to the construction industry.Based on this,this paper will transfer the deep learning technology to the construction safety management,through the collection of construction scene images,combined with BIM(building information system)Building information model(BIM)virtual scene,building construction scene data set,using UAV to obtain construction scene,loading it into the trained convolution neural network model of construction field,can realize the identification of construction scene and safety discrimination,and take a more scientific and rigorous management method as the basis for safety managers to manage the site.The main contents of this paper are as follows:(1)Through the track crossing theory and text analysis technology,the unsafe features of the construction scene are analyzed statistically,and the standard list of the unsafe features of the construction scene collected by UAV is formulated.(2)Using web crawler technology to collect a large number of construction site images,BIM virtual scene is introduced to establish the standard data set of general scene in construction field,and the standard data set of unsafe problems in general scene in construction field.(3)This paper introduces the deep learning technology into the field of construction safety management,designs the general scene recognition model according to the established data set,which can accurately identify the construction scene through model training;further draws the transfer learning technology in deep learning to the field of construction,innovatively migrates the deep learning model in other fields to the field of construction engineering,and constructs the deep learning model The general scene of the construction site is unsafe problem discrimination model,which can determine whether the construction scene is safe or not.(4)Combined with the above research results,this paper ends with the development of Python GUI Graphical interface,integrates the network model designed and constructed in this paper into the construction scene recognition and problem discovery system,which can realize the basic functions of recognition image display and result output classification.The paper has 92 pictures,33 tables and 100 references.