Recognition Of High-risk Scenarios In Building Construction Based On Image Semantics

Building construction risk inspection and recognition is a key link in construction safety management.Safety inspections based on manual experience not only have differences in the subjective experience of observers,but also lack the monitoring of real-time coverage.Therefore,the efficiency of the artificial risk identification method in the construction site risk investigation task is low.Computer vision technology has attracted more and more attention in the field of safety management.However,it is difficult to understand complex building construction scene information with the intelligent monitoring method of single object detection.In order to identify dangerous events from complex building construction scenes,this paper uses computer vision technology and ontology technology to understand the image semantics of building construction scenes,so as to automatically identify building construction risk scenes.The recognition of building construction risk scenarios based on computer vision technology and ontology technology involves the following steps.First,the sequence image containing the construction scene is exported from the construction site monitoring video.At the same time,the Single Shot Multibox Detector extracts scene elements of low-level semantic information from the image.The extracted image information includes the objects in the image and the detection rectangles of the objects coordinate.Secondly,an ontology semantic network is established within the scope of constructing a specific construction scene,and the ontological semantic language is used to convert the extracted low-level semantic information of the image into high-level semantics of event description.Next,the construction risk rules are converted to the ontology SWRL(Semantic Web Rule Language)rule base,and the high-level semantics of the image described by the ontology is verified by the Pellet inference engine to identify whether the construction scene in the surveillance video is high risk construction scene.Finally,taking the excavation construction scenario as an example,an automatic identification framework for risk scenarios is established,and the test results verify the effectiveness and practicability of the method.This study combines computer vision technology and natural language processing technology to carry out automatic inspection of construction site risks.This risk recognition method based on image semantics provides a new perspective for preventing the occurrence of specific construction accidents,and can be used to a certain extent to improve the efficiency of construction safety management.