Inside Chris Anderson's Open-Source Drone Factory 56
the_newsbeagle writes "The former editor of Wired is betting that the 21st century skies will be filled with drones, and not the military sort. His company, 3D Robotics, is building open-source UAVs for the civilian market, and expects its drones to catch on first in agriculture. As noted in an article about the company's grand ambitions: 'Farms are far from the city's madding crowds and so offer safe flying areas; also, the trend toward precision agriculture demands aerial monitoring of crops. Like traffic watching, it's a job tailor-made for a robot: dull, dirty, and dangerous.' Also, farmers apparently wouldn't need FAA approval for privately owned drones flying over their own property."
Summary is wrong (Score:3, Insightful)
Farmers most certainly would require FAA approval for using their own drones.
You can not use an ariel vehicle (ANY ariel vehicle) for commercial purposes in the US with a waiver or certificate of air worthiness.
Doing work for your farm would most certainly be commercial, even if you don't sell the product itself. Research alone can be commercial.
Exciting, but it will take a lot of work yet (Score:4, Insightful)
A lot of people are getting excited about what drone can do in agriculture. Folks on the diydrones forum, when they find out my brother and I actually are farmers, get all excited to try to solve problems for us. The problem is, it's going to take a lot of work to make drones useful in agriculture. I attended a presentation recently by an professor specializing in remote sensing and agriculture. She uses satellites, planes, and drones to try to get useful data from crops. It would be really handy to determine crop disease or monitor moisture use, etc.
Turns out, though, these are very hard problems, and small UAVs are actually making it harder in the short term. Here's why. A UAV map of a field, typically is done at low altitude, but stitching together thousands of high res images taken as the aircraft passes back and forth across the field in a pattern. Stitching is done using standard image algorithms that try to identify common pixels to line things up. The problem with this is that the very process of stitching the images changes the data. Is this pixel really this color of green, or did it get changed to fit in better (exposure adjusted)? Also the crop looks very different when you pass over it one way vs another way. For example a silk rug changes color if you view it from a different angle or rub your hand across the nap. This becomes a problem with UAV mapping because the resolution is so high, and the number of pixels is so great. With satellite imagery stitching doesn't really enter into it.
And once you get your image, what does it mean? I see some dark spots. Are these individual plants, rocks, dirt clumps, or shadows? Or is it horrible disease? And even if you can detect a difference in the crop's NDVI [wikipedia.org] pixel values, that does that mean? Is the plant just dry? Soil is naturally poorer? Or is disease. Sometimes disease shows up very clearly in an NDVI map taken from a drone. But in the end a human really has to walk the fields anyway, and take samples.
So the field (no pun intended) of UAV imagery is just getting started. I believe it will do cool things, but we have to be patient as we address the inherent problems with stitching, and also develops a means of understanding and exploring the data (google maps zooming for farmers' fields!).
For me the number one thing I'd like to get from UAV imagery would be accurate 3-d mapping of the topography for drainage purposes.
For right now, it's an expensive toy for some farmers to play with (UAV mapping and agronomy companies), and a project for researchers. And for me, UAVs are just a fun hobby.