Computer vision and hybrid reality for construction safety risks: A pilot study

Abstract

Construction sites are among the most hazardous venues. While most of the previous research has shed light on the human aspect, we propose to utilise the fast R-CNN object detection method to detect the construction hazard on sites and employ mixed reality to enable the artificial intelligence to detect the hazard. Fast region-based convolutional neural network object detection acquires expert knowledge to identify objects in the image. Unlike image classification, the complexity of object detection always implies an increase in complexity which demands solutions with regard to speed, accuracy and simplicity.