What I learned from scanning Alcatraz – Part 1

Article by Dr Jeremy Sofonia

Mapping notorious US island prison Alcatraz was no mean feat, requiring meticulous data capture across a rugged and complex 22.5-acre site and the processing of more than 100 scans during a 3 week period. Dr. Jeremy Sofonia, Emesent’s Senior Technical Specialist, talks about some of the tips and tricks he learned in this two part series – apart from wearing good shoes!

First up – the planning, which sets the foundation for success.

Part 1 – The Planning

Pete Kelsey and Dr Jeremy Sofonia at Alcatraz

Make a plan

When I first discussed the project with Pete Kelsey, the project lead, one of the first things we discussed was how important data management was going to be. While we knew there was going to be a lot of data coming through, I don’t think either of us fully anticipated the sheer volume.

However, we had already created all the file structures with folders for raw data, processed data, merged data, colored data and so on ready to go. So while the volume made things more complex,  with data at different stages of processing for example, all the pre-planning was still applicable, it was just blown out at scale. This highlights the importance of being rigorous about the initial setup because it’s easy to lose track otherwise.

Take versatile equipment

While our primary objective was to conduct handheld LiDAR mapping of the interiors, we were also tasked with capturing data in RGB ‘true-color’. This was easier said than done as many of the spaces on Alcatraz were not well lit –  or lit at all – which made the task all the more challenging. To capture better color, the engineers at Emesent designed a special lighting rig that snapped onto the Emesent Hovermap LiDAR unit and illuminated the field of view of the GoPro camera. 

For hard-to-reach places, the versatility of Hovermap meant it could also be mounted to an extension pole. For example, the very top of the lighthouse couldn’t be accessed without climbing a small ladder which was dangerous and because it was an active Coast Guard station would have required additional approvals. By attaching Hovermap to a pole, I could walk along the balustrade of the balcony to capture what I needed. Similarly, entry to a vacated building which no longer had step access would have meant climbing down a cliff face, clambering over rocks, and then climbing back up through a hole in the side of the building. Instead I was simply able to attach Hovermap to the pole and push it through the windows.

Ensure suitable processing power for your needs

One of the first things we did when we got to Alcatraz was set up a serious AMD ‘Threadripper’ with 99-cores and 512GB of RAM that was kindly donated to the project. It’s one thing to merge two or three data sets, but how do you work with 28 data sets from the cell house? Or the 30+ scans from the old apartments known as “Building 64”?

Emesent Aura uses the Unreal engine in the background, which allowed me to visualize a huge volume of data and do alignments with the merging tool without overloading the computer’s graphics card. That capability, as well as the power of the AMD machine, were critical in processing and visualizing such a large volume of data.

Visit emesent.com/alcatraz to hear the stories behind the scans.

File structure is important

When you are merging data and working with multiple computers it’s really important to have an agreed structure or path to work on and save the data. Appropriate file and naming convention will save time and trouble. We knew the names of the buildings and had already decided on our folder and scan naming conventions so that as we captured data, it could be organized, managed and processed efficiently.

On Alcatraz multiple people were capturing data in different areas so it was important to make sure everyone was on the same page and conventions matched up on everyone’s computers. An agreed structure allowed us to follow an identical path to go back and perform additional processes like merging and colorization. It may seem obvious but if you don’t get it right it can take a lot of time and effort to resolve. Essentially, the goal should be to implement a file structure and naming such that someone who didn’t directly participate can find, identify and work with the data from the site, down to the building, level and room.

Establish a tight survey control network

PCL, a major US construction company, was also on site and vital in setting up our survey control and relevant targets. Over a couple of days, they set up an extremely tight control with a reported error of  ±3-4 mm over the entire 22.57-acre island. This not only helped us ensure our data was georeferenced accurately but was also key in allowing data from the multiple parties involved to be overlaid and brought together as a single model.

Make a plan – plan to adapt

One of the things that made this project more difficult was the need to scan without interrupting the visitors that came to the island every day. Initially we planned to work very methodically, starting on the north end of the island and working our way south; starting with this building, then the next; scanning this floor, then this floor. But because we wanted to avoid the visitors it didn’t end up that way. In reality, we scanned where we could, when we could.

So, making a plan is important, but equally be prepared to adapt to changing or unexpected scenarios.


Ready to continue? You can find Part 2 of our series here.