Ontology-Based Semantic Modeling of Knowledge in Construction: Classification and Identification of Hazards Implied in Images

Abstract

Identifying potential hazards of construction project is a data-intensive process that involves various types of information such as site data, specifications, and engineering documents. How to effectively convert the information into a machine processable format for safety management is a challenging task. To address this problem, in this paper, combining the HowNet and specific taxonomies from the relevant construction specifications, a semantic modeling approach is developed for the proactive construction hazard identification from images. A semantic scoring system is then introduced for quantifying the similarities between images, via comparing their annotations with the construction hazard specification. Furthermore, an image processing framework is developed to semantically annotate site images and further automatically classify the images into the categories. In this way, the potential hazards implied in the images can be identified automatically. Examples are developed to demonstrate the feasibility of the approach. The outcomes of this study have offered an alternative method to enhance site safety management on site.