Integration of AR and deep learning-based image classification using CNN for construction project monitoring

With manual update, progress monitoring of a construction project is notably challenging due to its labor-intensive and time-consuming nature. The issue of manual update has been proposed to be alleviated by automatic update, but most existing AI-based progress tracking approaches are based on counting the elements finished. As a result, they cannot accurately reflect actual progress status to the project management. This research aims to develop a system that updates the progress of the construction project by effectively identifying both the construction category and operational stage. It is achieved by systematically integrating the techniques of Augmented Reality (AR) and deep learning-based image classification with convolutional neural networks (CNNs). The proposed scheme enables real-time data capture, facilitating comprehensive progress evaluation across multiple project locations. Its effectiveness is demonstrated using a case study of an interior finishing project for a building in Tamkang University, Taiwan. It reveals that the proposed system accurately identifies the construction category and operational stage based on the materials applied, enabling precise tracking of the progress. The results confirm that the proposed scheme significantly enhances the accuracy in construction progress monitoring, avoiding manual inspection and minimizing discrepancies between the planned and actual progress.