With photogrammetry technology being more accessible than ever, I’m sure there are a few beginners out there that are eager to give it a try. If you’re like me, you may have heard some of these tips before, but didn’t give them as much attention as you needed to. Here are some quick tips to help improve your photogrammetry:
Of all the tips in this list, this one is arguably the most important. When it comes to photography, video, and even photogrammetry, lighting is critical. At the end of the day, you’re just capturing light. So if the quality and/or quantity of light is not great, your results won’t be great either. So your objective for 3D scans is, counter to most photography, flat. Completely flat.
Try using big light sources like a large softbox or even a window to spread as much soft, diffused light onto your subject as you can. The fewer shadows, the better. If you can manage no shadows, awesome. A good technique for capturing people’s heads or small-to-medium sized objects is to evenly light your subject with two large keys from the left and right, and using a lazy suzan to rotate the object as you take photos. A swiveling chair or stool would work the same when photographing someone’s head.
Cover your angles
This one may seem obvious, but it took me a few tries to get it right. During my first few scans, I wanted to get in nice and close, hoping for some ultra high resolution textures as a result. Makes sense, right? If you get in nice and tight, grab all the high-res details, your final textured model will look stellar! Well, not quite… At least with the reconstruction algorithm I was using, photos that are ‘too different’ than the others (aka close-ups) were tossed out from the calculation, so they ended up not even contributing to the final.
That said, try to keep your entire subject in the frame at all times. This will definitely improve the chances of ALL your images being used to calculate the point cloud. Just a personal technique, I usually start with my low angles first, snap a photo, and move about 10 degrees to the right or left, snap the next and repeat. Once I make my first 360, I move up to the mid (head-on) angle, do a 360, then cover the high angles. It’s always better to have more images than you need, rather than coming up short during the model creation process.
This helps keep consistency between all your images. Most smartphones come with this feature already, but if not, there are several apps for both Android and iOS that will unlock your camera’s full potential. This prevents your camera from shifting focus, changing exposure, and altering white balance settings in between shots. That way, when you bring in all your photos, they’ll all be the same white balance, the same ISO, shutter speed, and aperture.
This goes hand-in-hand with the ‘shoot manual’ tip, but it’s definitely worth mentioning. If you happen to be shooting in low light, and you’re on a DSLR shooting prime or something like that, if you’re shooting wide open, you’re going to get a very shallow depth of field. It looks lovely, and I love the images from wide-open prime lenses, but this isn’t the application for it. When taking photos to use in photogrammetry, you need to keep your entire subject in focused. This is 2nd most important, right after light. This makes it easier for the computer to see exactly which angle you’re viewing the object from, and will likely result in fewer images being tossed out of the application. If you’re shy on light, don’t open the lens. Get more light in there to allow yourself to shoot f/10 or so, whatever it takes to get your entire subject in focus. Definitely helps you in the long run.
That’s all for now! If you have any tips of your own, feel free to share them in the comments!