Research On Reasoning And Automatic Recognition Of Workers’ Unsafe Behaviors Based On CenterNet

At present,the construction engineering safety accidents occur frequently in our country,and the unsafe behaviors of workers are widespread.Identifying and reducing workers’ unsafe behaviors as much as possible is conducive to improving the safety production situation of China’s construction industry.In recent years,the identification method of unsafe behavior of construction workers has become a research hotspot of scholars at home and abroad.However,most of the current research on unsafe behavior recognition of workers based on computer vision relies on artificial rules to determine the target,which has the shortcomings of weak generalization ability and low efficiency.In order to solve the above problems,this thesis proposes a reasoning and automatic recognition method for workers’ unsafe behaviors based on CenterNet.(1)On the basis of literature review and related theories and technologies,unsafe behaviors are defined and classified,and then the challenges and shortcomings of existing methods for recognition of unsafe behaviors of workers based on machine vision are analyzed.(2)An automatic object recognition method based on CenterNet is proposed.CenterNet can get the object and its center point through anchor free detection method,and get the feature vector of each object.The method combines the calculation of thermal diagram and the normalization of coordinate position,and establishes the model of the target object as the center point.The average accuracy of the method is 52.3% through experiments.(3)An unsafe behavior reasoning method based on Neo4 j is proposed.The spatial distance between objects in construction scene is calculated by machine vision method,the scene data is modeled by Neo4 j,and the unsafe behaviors of workers are identified by rule query mechanism.Finally,the effectiveness of the reasoning and automatic identification method of workers’ unsafe behaviors based on CenterNet is verified by experiments.It can automatically identify many kinds of unsafe behaviors of construction workers,which can reduce the occurrence of safety accidents on the construction site and promote the application of computer vision technology in construction safety management.