Our goal is to automate the process of construction performance monitoring by leveraging advances in computer vision, robotics and construction management. Construction monitoring can facilitate practitioners make project control decision efficiently and quickly. Current practice of construction projects is suffering due to coordination of labor and equipment, management of facilities and declining productivity, which is making a waste about 25-50% of the entire project. The automation process of construction performance monitoring can efficiently improve the productivity and reduce the cost by providing a characterize tool which enhance the progress and activity monitoring process. Contractors are already using hand-held cameras, video cameras collecting data to further analyze and organize. However, the cost and complexity of the collection, analysis and reporting operations result in sparse and infrequent monitoring, and thus some of the gains in efficiency are consumed by monitoring costs. Individual construction companies do not have the expertise to improve automation or the incentive to invest in the autonomous performance monitoring research needed to reduce construction costs nationally. We propose to improve frequency, detail, and applicability of construction monitoring by automating collection and analysis.
Automated Data collection
Manual collection procedures to take photos and install video cameras do not support the desired frequency or completeness of progress and activity monitoring. We propose to autonomously record videos for progress monitoring and place cameras for activity monitoring using aerial robots.
Construction projects benefit from hourly updates of how actual (as-built) progress compares to as-planned Building Information Models (BIM). Our prior work presents how to create a 3D as-built model from hand-held camera to compare with the as-planned model. This research aims to improve the completeness and reliability of the as-built models.
Vision-based quality control
Automated assessment of work-in-progress using 4D BIM as a priori and still images as as-built – with or without corresponding 3D point cloud models- has potential to significantly improve the efficiency of construction project quality controls. After creating precise and complete as-built models, this research purpose automated quality control method (steel structure, plumbness...) to improve the waste due to re-make in the construction industry.
Management of onsite construction benefits from near real-time analysis (every 3-10sec) of worker activities and equipment use. The collection and analysis is currently prohibitively expensive due to labor costs and is difficult to perform thoroughly due to frequent changes in the locations of major activities. We propose to develop automated video analysis techniques to partially or fully automate annotation of activity and resource use – construction activity analysis– and to inform placement of a camera network.