Handheld Photogrammetry Construction Interior Scan Use Case

Interior 3D scanning can be useful in a myriad of ways. In this article, I present a scan of a residential building construction site, with its foundation walls and roof almost complete. Further construction will require some visualizations and design according to “the fact.” “The fact” comes from a factual scan of the house, where any alterations from CAD drawings can be spotted and accounted for in further construction. I am sharing how I did such a scan in a couple of hours using a mirrorless camera.

Photogrammetry Equipment

I am using a Sony A7IV (a full-frame mirrorless digital camera) with a Viltrox 14 mm ultrawide-angle lens (a lens with a very wide field of view), which I tested in the previous article. The lens and camera combination is light and compact, so I will not be using any tripod or monopod. I would recommend using a brace if your setup is heavier or the scanned space is large. A monopod with an adjustable head would save strength in most cases without being too cumbersome.

The unfinished construction site is quite open to natural light, so I will not need to use any additional lighting equipment. If there were no abundance of natural light, this use case would become significantly more challenging. Ultimately, the only required equipment is your imaging device, complete with all its essential components.

Interior Shooting Patterns

My core interior shooting pattern is an inside-out perimeter, which means walking around the perimeter of the room while facing the opposite wall. I always make at least three such passes: the camera is positioned lower, facing more towards the ceiling; the camera is positioned square at the opposite wall; and the camera is positioned in an upper position, facing down to capture more of the floor. (A "pass" means a full walk along the perimeter, taking photos in a particular orientation.)

This provides a solid foundation for most average interior scans. The side overlap between photos must be at least 80 percent (overlap ensures proper photo alignment during processing). Measuring that precisely and adhering to it is practically impossible, so I visually estimate the step I should take between photos and repeat that as many times as necessary.

Each space must have at least one complete triple perimeter. After that, we can add additional data for specific areas or as connective tissue. In this particular case, I added connective tunnel-type passes between different areas, both in the forward and backward directions. With the camera facing forward and taking a photo at each step, as if going to another room. I did the same in both directions.

I added more orbital-type scans around key areas. One area is the understairs technical room (or rather, soon to be one) and around the corners where window bays meet walls. (An orbital scan means walking in a circle around an object or area and taking photos from different angles.) Finally, I took a walk around the house and on the roof to gather terrestrial exterior data (data of the building’s exterior captured from the ground), making the scan more complete. Unfortunately, I did not have the time to incorporate aerial data for the roof or the environment, which would have made everything look even better.

Camera Settings and RAW Conversion

I shot everything in portrait mode (vertical camera orientation), in RAW format, and with AUTO ISO enabled while in full manual M mode, using a shutter speed of 1/60th of a second. The aperture was set to f/4, wide open, with the Viltrox lens. That shutter speed is just safe enough when using an ultrawide lens, while the RAW image format allows me to maximise image quality, particularly in terms of dynamic range. The contrast level between the interior (no matter how well-lit) is huge when the sun peeks out through the window bays.

I kept my focus area in the center of the frame and refocused every time I changed my shooting direction, sometimes even more, just because I can. Autofocus on the Viltrox lens provides a safer feeling, ensuring that your subject (in this case, the wall in front) is always in focus. A manual focus lens with focus peaking would also work, but autofocus is one step more convenient.

In total, I got almost 2400 photos and converted them into JPEGs using Adobe Lightroom Classic. As always, nothing fancy: highlights down, shadows up, a bit of contrast via curves, “Remove Chromatic Aberration,” that is it.

Export all photos as JPEGs at 70% quality and import them into your favorite photogrammetry software. At this stage, I made a mistake that I always preach against; I did not filter the photos before processing. As a result, a few photos I took for fun made it into the set, and while it did not cause any issues, I suspect I lost a few minutes of processing time when the software tried to match these photos to the overall scene.

Data Processing and Referencing

I processed all photos using the default settings to create a 3D point cloud and a 3D mesh. The point cloud was the desired outcome, so I will be delivering that as the final result. In Pixpro, it took a few hours to process everything on a good desktop computer.

However, before I processed the 3D mesh, I cropped and cleaned up the point cloud slightly. I also oriented the point cloud upright using Pixpro’s manual referencing tools, just for easier navigation.

Referencing in Terrestrial Photogrammetry

Referencing was not required in this particular use case because point cloud registration will be done to a known framework (a reference system such as a blueprint or site plan). However, I still made a few basic measurements for proper scaling (resizing the model so dimensions match real life). I measured a few lengths that can be easily identified in the scene, such as window bays and stair steps. My technique is to lay down a measuring tape and take a picture with my phone as a reminder of where and what length each measurement is. I recommend making at least a couple of measurements in each room of the house.

Conclusion

This quick handheld scan demonstrates that interior photogrammetry does not need to be overly complicated. A lightweight camera, a reliable ultrawide lens, and a bit of structure in how you move through the space are enough to produce a solid point cloud that is good for real planning work. The shooting pattern is repeatable, the processing is straightforward, and the end result already gives a very clear picture of the current state of the construction. With a bit more time, I would have added aerial roof data and some extra exterior passes; however, even without those, the dataset turned out to be more than usable. A simple, fast scan like this is often all you need to bring a construction site “from drawings” into the real world.