I decided to run some older imagery I came across the other day through photoscan and came up with some pretty cool results. Above is the park behind 3d Robotics where I used to fly after work. Below is an aerial shot of the 3DR neighborhood.
And here’s a model of a rural area outside Jakarta on a hazy day made with images taken during a training flight.
If you’re working on aerial photography I encourage you to just get out and start taking pictures, you’ll be surprised what comes out when you’re just flying around randomly! Just don’t fly around former employers’ offices unless they’re into drones too ;-)
Comments
These are probably all about 50% overlap which means every point is seen from at least two angles but probably at least 4 since there's 40-60% overlap in the flight direction as well.
Thanks. How much overlap does photoscan require across images to give the results you show? 50%? I assume it needs some overlap to recover the depth relations.
All these pictures were taken with the camera pointing straight down with some small angles due to the roll of the aircraft fighting the wind. GoPros don't really work with 3d reconstruction in my experience... it depends on the software you're using to postprocess and whether or not you can calibrate the software to the fisheye lens.
Is it important to have the camera aimed at a 45 degree angle or straight down? Or can I just use the footage from my front-facing gopro off my FPV plane?
I was using the E382 from Event 38 (my company). I'm not sure about what cameras were used as these were so long ago, probably a Canon SX230 HS.
Thanks! I'll see if I can get around to making a KML...
Wow, very well done @Chris.
Could you make a small KML mosaic over 3DR em share it with us? it will be amazing!
Guto.
What were you using?