Construction Scene Hazard Source Identification Method Based On Machine Vision And Ontology

Construction industry safety accident is a serious and urgent problem to be solved.There are often unsafe behaviors of workers and dangerous working environment in the construction scene,which will directly lead to a lot of safety accidents on the construction site.The traditional inspection of construction site safety often depends on manual completion,which is difficult to meet the requirements of modern construction safety management.In recent years,detection of building construction scene based on computer vision has become an effective research method.However,the method has the limitation of semantic gap and can only extract visual relationship and does not solve the limitation of identifying unsafe conditions and behaviors by processing visual information.Given the above problems,this thesis carries out research on the whole process of "object detection-ontology construction-ontology reasoning" in the construction scene,in order to provide reference for the hazard source identification in the construction scene.The main work and research contents are as follows:(1)Relying on computer vision technology and deep learning technology,this thesis constructed the corresponding data set covering construction scene entities and selected YOLO v5 network for training.Then this thesis carried on the object detection to the construction site building entity,and extracted the construction site visual information;(2)Aiming at the limitation of semantic gap in the use of object detection,the ontology theory is introduced to construct the construction scene hazard source management ontology model,which realizes the integration of object detection and ontology model in the construction scene and solves the limitation of semantic gap;(3)To solve the limitation that existing studies do not process visual information to identify the construction hazard source,in this thesis,based on the proposed construction scene hazard source management ontology model,a targeted construction safety rule base is established.At the same time,this thesis introduces ontology reasoning to deduce the extracted visual information,the construction scene hazard source management ontology model and the construction safety rule base to realize the construction scene hazard source identification;(4)Based on practical cases,this thesis selects a common construction scene,with "workers did not wear safety helmet and walk under the crane arm" as an example.The rationality and application effect of the hazard source identification method for construction scene constructed in this paper are verified.This method can also be extended to other construction scenarios,which is helpful to the improvement of construction safety management level and the realization of intelligent management.In addition,the shortcomings of this method are summarized,and the future research directions are proposed.