Monitoring construction site situations with AI technologies

Activitiesin construction sites of buildings and or civil projects present significant risks to the workers’ safety due to unanticipated hazards and often neglected safety standards and protocols. Although technology has advanced significantly to minimize safety and hazards in construction sites, there is a pressing need to address how AI affordances affect cognitive and attentional abilities as indicators of engagement for information to deter safety and hazard violations or situations that might present associated risks to workers on construction sites. The research uses deep learning techniques to explore how users (e.g., workers and safety engineers) understand risks by presenting more relevant information, enabling a better implementation of safety standards. The study uses the researchers’ extended collection of images, documents, and designs from existing and previous construction projects as datasets to train deep-learning models. The approach employs a combination of semantic segmentation techniques and transformer-based models. The aim is to enable the detection and verification of any interference between current situations and associated risks during field inspections. The first step of the research identifies different typologies of potentially hazardous construction scenarios at the intersection of public areas and construction sites. The researchers will use an ontology built to classify and cluster-specific artifacts and scenes that present safety and hazards. By clearly describing the criteria and conditions in an ontology, the study builds a reference model to associate semantics to guide future safety and hazard machine-learning approaches.