Research On Unsafe Behavior Detection And Application Of Construction Workers Based On Image Recognition Technology

Construction workers as the main body of project production,their widespread unsafe behavior is the direct cause of frequent accidents.Therefore,it is necessary to take appropriate measures to reduce the unsafe behavior of construction workers.Although the current mechanism and influencing factors of unsafe behavior of construction workers has been discussed,the literature analysis finds that there is still a lack of in-depth thinking on macroscopic control.In view of this,paper combines the image recognition technology in the field of artificial intelligence to carry out the inspection and discussion of unsafe behavior of construction workers,designs the portal program and sets up an alarm system,in order to stop the unsafe behavior of construction workers,enhance the supervision ability of enterprises,depress or circumvent security incidents.Firstly,it sorts out the relevant content of construction workers’ unsafe behaviors,clearly defines the concept and scope of unsafe behaviors in research,elaborates the classification and mode theory of unsafe behaviors.On this basis,the author summarizes twenty-four unsafe behaviors of construction workers from the accident cases and the literature,exchanges and interviews with the construction management personnel to modify,then adjusts the remaining fourteen items which are divided into two categories four types: state and action,location,protective equipment,equipment operation and reaction processing.Secondly,establishes an algorithm service platform erection structure,compares the unsafe behaviors of the same item to verify the applicability of the model.Considering the number sample size of the database image collected by the simulation and cognitive intent,explains the commonality of the identification of the four types of unsafe behaviors and selects the four establishment models of the subordinates.By evaluating the model performance of image classification and object detection,it is related to the classification effect,detection target,image background coverage change,etc.At the same time,implements the picture verification,experience H5,online API and offline SDK,the probability values of the confidence are displayed.Finally,the application of model based on image recognition has been expanded and explored.Using the Python language and the PyQt5 environment to write a test program that visually expresses the recognition results,emerging the presence or absence of wearing a seat belt,is basically in line with the original intention.Planning a unsafe behavior identification solution and comprehensive disposal system,that establishes the identification model of unsafe behavior,then calls the interface to detect the safety of the behavior,and mixes highaltitude work,use of construction lifts,unbelted,climbing platform guardrail to reveal the image shunt to fit the model’s matching problem solving ideas,successfully obtained unsafe results in high-altitude operations.According to the alarm principle,an image recognition model is set up as a hub alarm system,due to the continuous improvement of accuracy and visualized before saturation,it has great potential for development.