One of the basic use cases for drones is to make amazing videos from unique vantage points. Most commercial drones produce high definition 4K video at 100Mbps or better. That makes for fantastic videos for marketing or other services! I was recently asked, “can that same high-quality video be used to create 3D models?” The short answer is yes…but don’t. Let me explain why.
Most structure-from-motion software like ContextCapture, Pix4D, or DroneDeploy can create a 3D model from a video. What the software does is chop the video into still images every few seconds. You, as the user, will define how far apart the images are taken. If you cover a lot of ground in the video, you may want it to take an image every second or two. If your video is relatively slow-moving, you may choose every 20 seconds or more. A video does give you a way of making sure that you collect data on your whole scene. However, there are several significant downsides to this approach.
1) First, still image quality is much better than a snip from a video. This fact alone almost guarantees that a video-based model will be significantly poorer than an image-based model.
2) With a still image, you have metadata embedded in it. Specifically, you’ll have GPS location, drone orientation, camera angle, and focal length. You lose all of this with the video. Without the metadata, your software may struggle to reconstruct the model well.
3) Getting consistent overlap and scene coverage is a lot harder with video, especially if you’re flying manually. Unless you program a pretty consistent flight pattern to cover your scene, at the perfect overlap and a constant speed, it’s very likely that when the software chops up the video, you’ll have too much data in some areas and not enough in others.
Bottom line, if a video is all you have, you can do it but throttle back your expectations. If you want a great model, use a good flight control app and solid image capturing strategies. Programing a control app and image capturing best practices are just some of the things we cover in the Applied Drone Technology course.