A novel integrated system utilizing a deep learning approach and the NIOSH lifting index to enhance construction safety

Abstract

Musculoskeletal disorders are a prevalent issue in the construction industry, largely due to the physical demands of object-lifting tasks. Effective ergonomic assessment is crucial for preventing these injuries and enhancing workplace safety. This study introduces a novel integrated system that utilizes advanced computational models to assess the NIOSH Lifting Index in real-time, offering a significant improvement over traditional ergonomic assessment methods. The system combines pose estimation, object detection, cycle counting, and Long ShortTerm Memory modeling to provide dynamic, real-time evaluations of lifting practices. The integrated approach allows for continuous monitoring and analysis of lifting tasks, providing immediate feedback that can be used to adjust working conditions proactively. This system was tested in a controlled environment, demonstrating high accuracy in predicting the lifting index and identifying ergonomic risks with impressive precision and recall metrics. The practical applications of this system in real-world settings suggest substantial benefits for improving safety standards and reducing the incidence of Musculoskeletal disorders on construction sites. The study also explores the challenges faced during the implementation of the system, including limitations related to pose estimation accuracy and the requirement for predefined object weights in detection processes. Finally, future research directions were also discussed.