A 1.144Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE
Imagine you are sitting at your office and would like to conduct a walk-through on your job site, but you are hundreds of miles away. Using a new D4AR–4 dimensional augmented reality–modeling technology, you can make a phone call to your construction site and ask your superintendents and field engineers to walk around the site, take photos and send them back to you. Then you can automatically reconstruct actual 3D point cloud model of the site using these photos and register the resulting point cloud model and photos with your Building Information Models. Using the D4AR models, you would be able to remotely walk through the site and study both the actual and expected status of your project. You can remotely monitor progress, productivity, safety and quality, perform geometrical measurements, or even analyze site logistics. Read on to learn more.
The Need for Automated and Visual Construction Progress Monitoring Technologies
Early detection of actual or potential schedule delay, cost overrun or performance deviation in field construction activities is critical to project management. It provides the opportunity to initiate remedial actions and minimize their impacts. This entails project managers to design, implement, and maintain a systematic approach for project monitoring to promptly identify, process and communicate discrepancies between as-built (actual) and as-planned performances as early as possible. Despite importance, systematic implementation of monitoring is challenging: (1) Current methods in the industry include manual as-built data collection and extensive plan data extraction from construction drawings, schedules and daily construction reports (Fig. 2-a), (2) due to extensive workload, monitoring is sometimes infrequent and performance may be measured with non-systematic metrics (Fig. 2-b); (3) Progress reports are visually complex. In these reports, monitoring metrics are visualized independently from one another; ultimately requiring more time to be spent in communicating status of a project in coordination meetings (Fig. 2-c). There is a need for an application to collect site data easily, process the information automatically, and report back in a format useful for all project participants.
Daily Construction Photo Collections and Building Information Models as Emerging Sources of Information for Automated and Visual Progress Monitoring
Nowadays, cheap and high resolution digital cameras, low cost memory and increasing bandwidth capacity enable capturing and sharing of daily construction photographs on a truly massive scale. In many building and infrastructure construction projects, hundreds of photos are being collected by professionals on a daily basis. Such a large and diverse set of imagery allows a site to be fully observed from almost every conceivable viewing position and angle during construction of a project.
In the meantime, Building Information Models (BIM) are also increasingly turning into binding components of construction contracts. If linked with project schedules, BIM can form chronological models to analyze expected performance of a project and revise construction schedule accordingly. 4D (3D + time) BIM can also serve as a powerful baseline for tracking and visualization of performance discrepancies.
Using site imagery and BIM as emerging sources of information, recent research at Virginia Tech and formerly at University of Illinois has resulted in a new modeling technique that uses such common photos to visualize and automatically track construction progress in four dimensions, offering construction professionals a new, low-cost way to monitor projects. To develop the system that generates these D4AR–4D Augmented Reality models, Prof. Mani Golparvar-Fard, Assistant Professor of Civil Engineering at Virginia Tech worked with Prof. Feniosky Pea-Mora, now Dean of Engineering and Applied Science at Columbia University, and Prof. Silvio Savarese, Assistant Professor of the Department of Electrical Engineering at the University of Michigan.
Using digital photos of modest resolution, the system constructs dense, three-dimensional point cloud models of the construction site, automatically computing each photo’s viewpoint in 3D. Using different photo collections taken over time, it generates four-dimensional (3D plus time) point cloud models. Finally, BIM are linked with construction schedules and superimposed with the point cloud models. The results are D4AR models that visualize actual and expected models together and automatically color code progress deviations based on a simple traffic light metaphor. Fig. 3 a, b and c show a case study where using 160 2 megapixel images captured with a digital consumer camera, the point cloud of the construction site is reconstructed and the photos are geo-registered in 3D (Fig. 3-d). In this case, there is no need to use GPS or wireless for tracking the camera. The location and orientation of the camera are automatically calculated using the visual content of the image. Fig. 3-e shows the 3D BIM which is superimposed over the point cloud model and finally Fig. 3-f is visualizing a column which is behind schedule.
These D4AR models (See Fig. 4) allow remote and easy virtual walk through on the construction scene, facilitate remote construction control decision making, minimize the time required to discuss the as-built scene through quick and intuitive access to actual construction information and can significantly cut in travel time and cost for project executives, architects and owners. Using the D4AR models, construction professionals can remotely access a project, visualize integrated as-built and planned site and assess progress, productivity, safety, quality as well as site logistics. Unlike laser scanning or time-lapse photography technologies, this technology which only uses casually collected photos from any type of digital consumer camera is able to use existing information to reconstruct 3D and 4D point cloud models of the site, visualize both expected and actual status of a project, and facilitate detection of performance deviations without adding the burden or cost of explicit data collection on project management.
As they say, "A picture is worth a thousand words, so you can imagine the value of hundreds of pictures combined together".
Mani Golparvar-Fard is currently an Assistant Professor of Civil Engineering at Virginia Tech. He is also director of Real-time and Automated Monitoring and Control (RAAMAC) research lab (www.raamac.cee.vt.edu) which focuses on creating and developing systems that can automatically track construction performance metrics using site digital photos and video streams as well as building information models. As an entrepreneur, along with Feniosky Pea-Mora (Dean of School of Engineering and Applied Science at Columbia University) and Silvio Savarese (Assistant Professor of Electrical Engineering at University of Michigan), he is launching a new startup company (Vision Construction Monitoring LLC) to offer the D4AR modeling technology to the construction industry.
A 1.144Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE