Computer Vision in Construction: An Introduction and 5 Use Cases

The construction industry is one of the world’s largest and most essential industries, comprising 13% of global GDP and employing 7% of the world’s working population, according to a report by McKinsey.

But despite its elite position on the economic stage, construction faces many issues. Its productivity growth rate has been lagging behind at only 1% compared to the global industry average of 2.8% over the past twenty years, and it’s also the most dangerous industry in the United States, accounting for more workplace fatalities than any other industry.

New solutions are desperately needed to improve business and to protect worker health and safety.

Thankfully, developments in a field of artificial intelligence called computer vision can help mitigate these problems. What’s more: computer vision solutions are often easy to implement, meaning construction companies can integrate these cutting-edge technologies into their workflows to immediately improve efficiency and safety on the jobsite.

What Is Computer Vision in Construction?

Computer vision (CV) is a field of artificial intelligence (AI) that focuses on the ability of computers to derive meaning from visual inputs, like photos and videos. In other words, it’s an area of research that tries to figure out how to make computers see and process the world like we do. 

For example, how can we get a computer to tell the difference between a dog and a cat? To recognize a face? To identify cancerous moles or to catch inefficiencies in traffic flows? These are the types of problems that CV scientists and companies work on.

Computer vision in construction is, then, simply the application of computer vision technology to the construction industry. It’s most commonly used for:

• Progress monitoring and cost control
• Productivity analysis
• Safety management
• Quality control
  

How Does Computer Vision in Construction Work?

Computer vision is a complex topic that requires years of study to fully comprehend, but it’s important to have a basic level of understanding if you’re considering integrating it into your company’s workflow. 

There are three basic steps to the process: data acquisition, data processing (image processing), and semantic understanding. 

First, a computer needs to be able to see — or in technical terms, to acquire visual data. To do so, computers can use 2D devices (surveillance cameras, cameras, smartphones, etc.), 3D devices (stereo and RGB-D cameras like Microsoft Kinect), or 3D point clouds (laser scanners) to capture real-time data. Commercial computer vision systems, like Safari AI, offer multiple options, allowing you to either use the equipment you already have or to use specially-designed equipment.

Once the computer has a set of eyes, it needs to process the data. That means the AI will use various algorithms, such as convolutional neural networks, to translate the images into a format that it can understand. In construction, this often means creating a “skeleton” for construction workers and machinery, like so:

‍

After that, the computer starts to make sense of the scene. It’s at this point that it uses deep learning to derive actionable insights through scene reconstruction, pose estimation, object detection, and activity recognition. 

In other words, it “looks” at the simplified models and figures out how they relate to each other to understand what’s happening on the construction site: it recognizes skeletons that fit a human form as people, skeletons that have a mechanical form as machinery, flat areas as work areas, etc.

For example, the computer may start by dividing what it’s seeing into two segments, such as a digging and a loading area:

Then, if a roller is detected in the loading area, even though it’s not supposed to be there, it determines that there is a risk of collision, and automatically alerts a worker, supervisor, or project manager, who can move the roller and fix the problem:

5 Use Cases of Computer Vision in Construction

Currently, computer vision is used to:

• Improve workplace safety
• Analyze productivity and improve efficiency
• Monitor construction progression and control costs
• Refine quality control
• Make better planning decisions

Improve Workplace Safety

Construction is the most dangerous industry in terms of workplace fatalities. In 2020 alone, 1,034 construction workers died in the US due to on-site injuries. The four most common fatal injuries, aptly dubbed “The Fatal Four,” are:

• Falls
• Struck-by (ex.: getting hit by an excavator)
• Caught-in/between (ex.: getting stuck in a trench and being crushed by machinery)
• Electrocutions

Construction companies can use computer vision to automate safety monitoring and detect when workers are putting themselves in dangerous situations. 

For example, computer vision software can detect when workers aren’t wearing proper personal protective equipment (PPE), such as hard hats, safety glasses, gloves, and vests. If the software notices that someone isn’t wearing a vest, which increases the risk of struck-by injuries due to decreased visibility, it can automatically send an alert to a supervisor who can then take action. 

Similarly, CV can recognize when workers are entering unsafe areas or standing under loads to reduce caught-in/between injuries. It can also detect when speed limits are being exceeded or when adequate fall protection is absent in an area. CV can monitor for ergonomic injuries due to repetitive movements as well.

‍

Watch a video on Computer Vision and Construction Safety

‍

Analyze Productivity and Improve Efficiency

Researchers at the University of Delaware believe that construction’s productivity is lagging behind other industries in part because typical on-site management processes do not identify productivity issues quickly enough. This means that supervisors are unable to take corrective actions effectively. 

To help solve this issue, they designed a computer vision system that identified when workers were idle. The software would automatically notify project managers so that they could ask them to continue working. 

This is just one of many ways that computer vision can be used on construction sites to ensure jobs are progressing efficiently, and it shows just how powerful the technology can be. Other applications include:

• Automatically calculating shift production using time-lapsed images
• Monitoring the availability of resources to reduce the amount of time workers spend waiting for materials to arrive
• Generating heat maps to identify productivity bottlenecks, such as over-crowded spots or long distances between material storage and construction areas
• Monitoring pedestrian and vehicular traffic to find the best time for closures

Monitor Construction Progression and Control Costs

In addition to detecting people and machinery, computer vision can also detect other important construction objects, like columns, electrical outlets, and drywall, as well as recognize different materials. This allows CV systems to monitor the progress of construction sites.

For example, if a camera is pointed at a concrete wall, it can calculate how much of the wall has been covered with insulation and enter that number into a database. If cameras are set up facing several different task areas, project managers can instantly get accurate data regarding task completion without having to visit several areas on the jobsite and eyeball the progress themselves. 

When CV systems are used together with Building Information Modeling (BIM) tools, the computer can detect differences between what it’s seeing in real time and the BIM data. It can then quantify the difference and accurately report progress. 

Plus, the object detection and material classification capabilities can help ensure that the project is staying not only on schedule but on budget too. 

Refine Quality Control

When it comes to quality control, computer vision tools can help in two ways: making sure that tasks are completed properly and that materials are in good condition. 

Unlike human vision, computer vision typically breaks down scenes into geometrical objects, like polygons, to help make sense of the scene. This makes CV systems incredibly adept at detecting small changes in geometry that could affect build quality. 

For example, a CV system can easily detect if a column isn’t perfectly straight or square even if it might look perfectly fine to a human. That’s because computers inherently calculate precise angle degrees and measurements instead of just going by what looks right — if one of those measurements isn’t right, it will notice immediately.

CV systems can also calculate an object’s flatness, whether reinforcement bars have been positioned properly, and whether any materials have been degraded by the weather. 

Make Better Planning Decisions

Most computer vision applications focus on on-site applications, but CV can also be very effective before a project begins. Developers and construction companies can use computer vision to determine the best locations for new developments and the best times to do construction.

Commercial solutions, like Safari AI, allow developers and construction companies to:

• Monitor pedestrian and vehicular traffic to determine the best time for road closures
• Count vehicles to figure out how a new development will affect an area
• Assess parking space and delivery/fire-lane availability in real-time

Computer Vision in Construction: Key Takeaways

Computer vision is an emerging field of artificial intelligence that utilizes the power of modern computing technology to process visual data for many different applications. In construction, computer vision is being used to improve worker safety, improve productivity, monitor progress, keep track of costs, enhance quality control through automation, and make better planning decisions.

Ready to improve your construction workflow? Learn more about how Safari AI uses people counting to help plan road closures.  

‍

‍

References

Xu, Shuyuan, Jun Wang, Wenchi Shou, Tuan Ngo, Abdul-Manan Sadick, and Xiangyu Wang. "Computer Vision Techniques in Construction: A Critical Review." Archives of Computational Methods in Engineering 28.5 (2020): 3383-3397.

Kadoura, A, and E P Small. "Tracking Productivity in Real-time Using Computer Vision." IOP Conference Series: Materials Science and Engineering 1218.1 (2022)7.

Occupational Safety and Health Administration [OSHA], OSHA Quick Card Top Four Construction Hazards, https://www.osha.gov/sites/default/files/publications/construction_hazards_qc.pdf.

Ranaweera, Kamal, Janaka Ruwanpura, and Siri Fernando. "Automated Real-Time Monitoring System to Measure Shift Production of Tunnel Construction Projects." Journal of Computing in Civil Engineering 27.1 (2013): 68-77.

Rezazadeh Azar, Ehsan, and Brenda McCabe. "Automated Visual Recognition of Dump Trucks in Construction Videos." Journal of Computing in Civil Engineering 26.6 (2012): 769-781.