Experimental platform for industrial worker behavior analysis based on computer vision
[Objective]Computer vision technology has been widely used and has achieved satisfactory recognition results in industrial safety.However,most studies mainly focus on the static identification of a personnel's protective wear or noncontinuous unsafe behaviors,and they fail to integrate and apply artificial intelligence technology with industrial safety control standards.This paper develops an experimental platform to analyze the behavior of industrial personnel using computer vision technology.The platform is integrated with industrial safety control standards for the practical teaching of undergraduate and postgraduate students.[Methods]The function structure of the experimental platform includes the resource,algorithm,and application layers.The resource layer relies on the engineering training center,in which multichannel data acquisition devices are employed to collect data on the work environment,worker operations,and labor protection appliances.The algorithm layer implements the identification and analysis of an industrial personnel's unsafe behaviors using computer vision algorithms involving human body key point detection and behavior recognition.The application layer provides the platform with diverse functions,including the analysis of a personnel's unsafe behaviors and behavioral data archiving.Moreover,intelligent control methods facilitate the integration with the algorithm layer,enabling the customization of application services according to the specific requirements at the algorithm level.This process also ensures the monitoring of enterprise-wide production processes for safety purposes.The industrial worker behavior analysis experimental platform is designed and built using PyQt5 and incorporates modules for human-computer interaction,data input,personnel behavior analysis,and data storage.The human-computer interaction module implements the platform's function control and data visualization,facilitating interaction between various sectors through function buttons.Further,the interaction module generates keyframe images or videos in the designated display area while concluding the personnel behavior analysis and concurrently outputs text descriptions of the analysis results through the analysis result display area.The data input module collects and transmits behavioral data.The personnel behavior analysis module employs various algorithmic models to analyze and process the video data from the data input module according to the specific requirements of different industrial scenarios.The data storage module manages the storage and retrieval of identity information for industrial personnel,behavioral video data,behavioral analysis results,and corresponding keyframe images.[Results]In the verification phase of the experimental platform,the operational behavior analysis of the personnel was performed.The experimental results demonstrate that the platform supports various safe work behavior analysis algorithms and enables automated identification of an industrial personnel's work behavior throughout the video capture process,according to industry safety control standards.This capability enables the effective assessment of diverse worker operations.[Conclusions]The proposed experimental platform can help enhance the intelligent image processing skills of undergraduate and postgraduate students.Moreover,it can serve as a reference for designing related industrial safety monitoring systems,improving the level of industrial safety,and offering potential practical applications.
computer visionindustrial safetybehavior analysishuman key point detection
万方数据
维普
