Getting first impressions of UAV/Drone forest images

Starting an unmanned aerial vehicle (UAV) survey is a shot into the dark. You have to buy the drone and equipment such as cameras , fill out a hundred of forms to get flight permission, spend another hundred hours in front of the flight simulator to learn how to fly the thing. The drone also requires assembly and, on top of that, establishment of a work flow for the image processing before you even get your first sight on the data. In the end, the first drive to the research area can end in an unexpected, but very solid, barrier.

The first flight is a big adventure with plenty of technical issues, but after a while you become familiar with the routine and the most obvious mistakes, like forgotten cables and empty camera batteries, appear less often. This is when things start to become fun. 

Taking a look to the first images is overwhelming. All the investment and work in using a high-performance copter with a great full size-sensor camera, instead of a ready to fly consumer device, was worth it. The images have an astonishing quality.

However, beautiful images in nice quality is only the start. If we want to say something about forest structures, we need to compile these images to a 3D model using a structure from motion (SfM) work flow. We used Agisoft Photoscan to generate a 3D point cloud and a 2.5D digital surface model. To be honest, I was really impressed by the amount of details we get, even below the canopy cover.

Now the real work can start, flying above more than a hundred additional research plots and analyzing the data for forest structures.

by Julian Frey (A1)