System for monitoring and preventing occupational accidents for the construction sector

Abstract

Introduction. The present study proposes the design of an occupational accident prevention system, which is made up of two elements: the first, a psychometric component, in which a scale was validated to measure the attitude of workers regarding the use of equipment of personal protection (PPE). Methodology. For this, an instrument was developed with 16 items on the Likert scale, obtaining reliability in Cronbach's alpha of α = 0.806, after applying exploratory factor analysis, three factors were obtained, the first factor retains a 26% variance, the second a 25.9% variance, and the third retains 14.9% of the total variance; The three factors retain a total variance of 66.8%.  The second component uses a Deep Learning applied to computer vision for the creation of a detector of the use of PPE that will contribute to the prevention of risks and occupational accidents. The detector works with videos and images acquired in constructions applying the YOLO algorithm that segments areas of interest and detects whether the worker is wearing PPE. In the latter case, an alert record is also obtained. Results. The system with the components described is applied before and during the execution of work. In this sense, prior to the hiring of personnel, the psychometric measurement instrument is applied to obtain the worker's profile regarding the use of PPE. Also, the respective monitoring will be conducted with the application of computer vision. With this information collected, a database is established that will monitor the workers according to the psychometric profile and the absences registered with the artificial vision to obtain later statistical projections that allow making decisions regarding the motivation or additional training for the workers in the use of PPE.