No. 1 Mistake You Are Making in Photogrammetry Right Now

As photogrammetry software developers, we need to troubleshoot multiple projects from our clients every week. The number one mistake, especially when making measuring projects, is using redundant photos. These photos can appear for a few reasons, which I will describe, and they can be quickly dealt with. Technically, it is a minor and avoidable issue, but we see it as the number one reason why projects fail altogether or succeed in processing but provide skewed results—which is the worst effect.

Main Data Issues 

There are three categories of harmful photo data in photogrammetry projects:

• Redundant data – photos that do not provide any unique data to the 3D scene.

• Irrelevant data – photos not of the subject that the software has to needlessly process.

• Low-quality data – technically failed photos are not up to standard in providing helpful information.

Low-Quality Photos 

Let me analyze these in reverse order. Low-quality data consists of blurry photos, which we can avoid by monitoring and adjusting the exposure (photo exposure consists of shutter speed, ISO, and aperture settings, which determine how bright the photo is). The quality of an image can be subjective, but that is not the case in photogrammetry. In photogrammetry, a good photo is one where you can see the subject you are scanning without distortions.

So, no blurriness (in focus and no motion blur), no odd coloring (proper white balance, usually), and no under- or overexposure (photo being too bright or too dark). If a photo meets these criteria, you can see every detail of your subject, meaning you have a high-quality image. Low-quality photos often happen when shooting conditions lack light, such as on a heavily overcast winter day or at dusk before sunset. Watch your exposure settings if you do not need sunglasses to be outside.

Irrelevant Photos

Irrelevant data usually consists of photos that were added to a project by accident. Such images can vary from something reasonable, like photos taken during takeoff or landing, to something ridiculous, like accidentally including pictures from a family outing yesterday with a backyard barbecue. Yes, it happens, and the best way to prevent it is to quickly inspect the photos in the photo manager before clicking OK to upload them into the software. Pixpro photogrammetry software includes this step for this very reason.

Redundant Photos

Lastly, and most importantly—redundant data. Redundant data directly results from inferior flight planning or overzealous manual shooting. The concept is dead simple—photos taken from the same position in space. Suppose you do not move between consecutive shots in photogrammetry. In that case, you are gathering redundant data, which can harm the project overall, as the software will try to use it for 3D reconstruction. And crucially—this applies to pivoting and tilting the camera. Moving only on an axis and not in space does not count as shooting from a different position.

In some cases, this can be hard to track. With programmable enterprise drones, there is an option for interval shooting, which does not stop even if the drone stops, thus producing many photos from the same spot. During manual photogrammetry, the same applies, but there is also the factor of user error—shooting photos without moving just because. Again, no two images should be taken without moving in space with your drone or camera.

Pixpro Waypoints Tips

There is also our solution service. Pixpro Waypoints works in such a manner that redundant photos should not be a huge issue. When using Pixpro Waypoints, the user is responsible for triggering photo capture for each flight plan. So there is one rule—always trigger the capture as the drone moves in the planned trajectory. That is it. Do not be tempted to trigger the capture earlier as the drone gets into position. Let the drone complete its path, and then, when it is slowing down at the last waypoint, end the timed shot. The same applies to all orbitals and grids.

Fix It in Post

Even if the capture resulted in multiple photos taken in the same spot, the fix is quite apparent—delete the repeating images. Pixpro Photogrammetry software has a photo manager, the first step in all processing workflows. Scroll through the photos and delete each one that is too similar to the previous one. By "too similar," I mean there is barely any movement between them, or, more precisely, the frame did not change by about 1/5th in the direction of movement.

Usually, we do not need to be precise in following the 80 percent overlap because these photos happen in clusters. So leave one photo in that position, and you are golden. Choose the best one visually and discard the rest. This is only a matter of doing additional photo inspection—maybe just a few minutes of work. And if you find these after the processing is complete - you can rerun the workflow again 

Conclusion 

Do not skimp on the photo inspection step. After doing hundreds of scans, I still do this, and I occasionally find a bad photo here and there. We have also encountered many projects where dozens of photos were taken in one spot, throwing off the scale of the entire project. And to be honest, we have tested this not only with Pixpro. The effect can be pretty substantial—scenes can shrink or expand by even a factor of two. So check your photos, delete what is unnecessary, and you may solve all issues regarding measurements.