Ontological inference process using AI-based object recognition for hazard awareness in construction sites

Abstract

Research on automating risk situation recognition using AI is being actively conducted. However, hazard situation recognition using AI has a limitation in securing a large amount of data for training of all hazard situations.This study proposes an approach that combines the conventional AI image recognition technology and relation-based reasoning to overcome the limitations of the method of sufficiently training each image regarding various hazard situations. To validate the proposed process, we constructed ontology by defining the relations between construction site objects and working situations, safety situations required for each working situation, and hazard situations based on the case of the work using mobile scaffolding. This approach will enhance the efficiency of using AI by inferring the current working situation based on the relations between recognized objects and determining whether it is a safe situation based on the inference on the standard safety situation for the corresponding working situation.