Visual Intelligence — unprecedented insights for the construction industry

Authors: Chandler McCormack is VP Building Solutions Field, Insights, at Hexagon's Geosystems division and CEO of OxBlue, part of Hexagon

For decades, leaders across all industries have used photos to see, understand and act upon insights appropriately. However, todays AI-enhanced construction cameras provide a level of insight and awareness that was previously unattainable.

Yet the information within an image isn’t always so evident. While technology has rapidly evolved, many data points within a photo remain untouched and unexplored.

While many industries have focused on facial recognition through image analytics, and artificial intelligence systems can now generate images with just a few prompts – actually unlocking the mountain of information in a photo is still a challenge.

  • How do you understand what’s happening in one snapshot of time?
  • How do you reveal the history of that time?
  • How do you begin to see the circumstances behind the photo?

In the construction industry, photo documentation is a natural and deeply embedded part of the job. General contractors, owners and developers all have a need for visibility so they can take the right actions at the right time. But what was once just a form ofrecord keeping is now a crucial part of managing, monitoring and marketing jobsite success.

Clear visuals are a timeless solution

Before high-speed internet, plugging into the jobsite was exceptionally limited. Webcam views could only offer a 320 x 240 video stream but this view was, and still remains, essential. When construction begins, sharing information doesn’t stop with just the team on site. G.C.s must coordinate with architects and engineers. Developers need to set schedule expectations for investors. Owners need to ensure their investment is secure.

In the late 1990s, the challenge for visibility was about sharing live visuals clearly and quickly.
Dial-up connections reigned supreme, a dreaded option that clogged phone lines and took substantial time to transmit. Images would load pixel by pixel, line by line, excruciatingly slowly. Construction camera providers had to build out servers and call lines so that instead of turning off phone lines for a few hours, construction teams could find a camera service provider who could buffer the data through a 1-800 line.

When the first construction cameras came online in the early 2000s, it was an attempt to deliver real-time information to jobsite stakeholders, fast. It was a way in which to finally “keep the record straight.” It was also the beginning of automating documentation. Images included metadata, a key piece of information that sourced the date and time of a photo’s creation, automatically.
Construction professionals could review jobsite changes and filter those changes by date and time. Now, photos could be tied together to create time-lapses of progress. In 30 seconds, clients could see the history of their jobsite in detail.

But metadata was easy to change or remove entirely. Simply saving an image from one camera to another computer desktop could wipe out the photo’s digital history. The data tied to images still lacked significant context; what was the weather like throughout the day? What else could be pulled out of an image and used to make decisions? What did images contain that wasn’t easy to see on the surface?

By 2012, construction cameras grew from a silent sentry on the jobsite to a key piece of a functional project management system. Cameras:

  • Made it easy to access one or more projects from anywhere and on any device
  • Delivered unbiased, unfiltered documentation to mitigate risk and rework
  • Shared real-time project status to all team members

As computer and cellular systems advanced, so did the camera. Technology began to help unravel the data points within every image.
Weather data and precipitation levels were time-stamped and marked appropriately. Optical zoom became a powerful means of exploring detail on a live camera feed. Simply being able to click into a project and understand what was happening over the last day, week, month or entire lifecycle of the work, was game-changing. Changes within a camera were driven by megapixel optimisation, video stabilisation, advanced zoom and robotic control from all devices.

What was important by the late 2010s was maintaining sharp, ultra-clear insight into the jobsite’s progress. Camera providers began to find ways to fix literal visibility issues, such as if there was dirt on the lens, or if the view had become tilted. It was important to deliver construction cameras that clients could “set and forget”, focusing only on seeing the real-time details of a job come to life on their screens.

Then the pandemic hit.

Artificial intelligence adds breakthrough management control

Visibility shifted from being about understanding day-to-day construction progress to understanding what was happening across the country. At OxBlue, we stood on the vanguard of using visual data to see reality during an unprecedented time. We started our journey into using artificial intelligence to manage our own internal support service. It was important to give our clients peace of mind that camera management was under control by their trusted provider.

A construction camera typically captures 25,000 images over the year that it is positioned on site. Those image collections partnered with artificial intelligence turned the construction camera from a device for decision-making now into a tool offering invaluable context for reflecting and improving in the future.

The camera in addition to the AI solution created an invisible force; a level of engagement akin to having the whole team on-site at all times. Real-time vision removed the need for waiting for quarterly reports to assess the impact of the pandemic on the construction industry; the images, bolstered by AI, showed where there were slowdowns and where activity was spiking.

Artificial intelligence began to unravel the context within jobsite images and videos. Activity data came first, then safety information regarding hard hats and high- visibility vests. A more complete picture started to emerge. Now, clients could see the correlation between labour hours and weather hazards. AI could “review” images and pinpoint the objects within them. Instead of sifting through months and years worth of data to understand how often equipment was being used, or which teams were more operationally efficient – AI could serve information on a platter, in an easy-to-read format.

This type of photo-review by AI is still relatively new. While Google images came to life in 2001 due to Jennifer Lopez’s infamous green dress – the search giant can’t use those images to tell you how many people were in attendance to see her, or how long they stayed for her red- carpet walk. Other, recent innovations, like DALL-E, are focused on serving up information based on a user’s query, rather than compiling that photo information on your behalf.
Construction cameras with artificial intelligence can now provide a level of insight and awareness that is impossible to achieve single-handedly. With artificial intelligence reviewing images, it becomes easier to answer big questions:
  • What factors are driving delays on the job?
  • Where are resources being underutilised?
  • Are my suppliers fulfilling their promises?
  • When and where did our equipment arrive?
  • Did last weekend’s storm impact progress?
  • Are things going according to plan?


The evolution of camera images is the evolution of efficency

AI has revealed data that allows general contractors, owners and developers to monitor the day-to-day with ease, but more importantly, it provides context and insight.
Now, images aren’t a matter of seeing a snapshot of time, but part of weaving together the situation and circumstances behind the photo. Whereas before, photos could be used to check in on progress, they are now a source for driving change. Visibility is valuable on its own, but understanding alongside that visibility is what will determine the future.

By consistently measuring site actions and activity over time, industry professionals have access to deep insight that will inform future planning. Cameras can now be a measure of “stress testing” for the unexpected.

What does it mean now that you can see that, while team A can deliver a final product in record time, they also have more safety violations than team B which consistently delivers a week past deadline? How different is life on the jobsite knowing that equipment rentals only have a utilisation of
30% over a span of a few days, instead of the assumption that they were needed 100% of the time? Understanding what key conditions look like, or what consistently gets in the way, means knowing how fast your team can turn things around, or what needs to be avoided in the future. These data points can all be essential diagnostic tools for what changes need to be made.

We've only scratched the surface

AI is still in its infancy. But even so, what’s happening today is beyond promising. Construction teams can now know if there is low activity on site for two days in a row without checking. They can now monitor if safety concerns have passed a certain threshold and require attention. AI can even look for motion and send an alert to a security team to dispatch authorities.
The full potential of AI is yet to be released, and yet cameras with this technology can already:

  • Quickly verify jobsite activity, equipment delivery and material use
  • Report findings that highlight when something on site needs your attention
  • Support compliance of evolving safety, health and regulatory standards

While there is more to come, we know one thing for certain. This consistent collection of data will help us all understand how to create consistent results, and do so more efficiently than ever before.

Contact Leica Geosystems

Find your Leica Geosystems contact for sales, support and technical service.
Find your Leica Geosystems contact for sales, support and technical service.