First test of passive point cloud generation from my new Upair2 3D+4K drone. Point cloud image is generated from a single frame of my neighborhood on the first flight of the drone.

I purchased an off-the-shelf 3D capable drone in order to test out ideas and get some practice flying before I try to build one myself with my GS-Beta . Luckily I found a manufacturer that makes a dual camera ready-to-fly drone: The UpAir2 3D+4K. It was about $500 from amazon and so far I’m pretty happy with it.

3D point cloud generated from a single 3D image capture from the Upair2 3D+4K after camera calibration shown below. Single threaded MATLAB code using sparse background grid and edge based features with windowed normalized cross correlator and subpixel (~0.2px) correspondence registration error. Run time about 10 minutes.

For just getting started and learning to fly a drone with 3D FPV the upair works just fine. But I’m also interested in generating 3D point clouds and scientific instrumentation, and the upair does not have synchronized dual cameras and utilizes rolling shutter on the 2 x 1080p cameras used for FPV / 3D capture. The baseline between the cameras is also fairly small, but the data is still pretty good and fun to play with. Image captures do not have embedded GPS or IMU data stream either. (I plan on building my own payload and platform soon that has all the bells and whistles I want.)