Produce costs farmers much more money to grow on a per-acre basis, since they need so much more TLC. With produce, quality is everything to the farmers, and there are many more opportunities during the growing season to make changes that will improve the end result. We've all put bruised produce back on the shelf at the grocery store, and for every one of those pieces, there are several the didn't make the cut!
Hi Hacker News. This is Nikhil Vadhavkar. I'm a cofounder of Raptor Maps with Eddie Obropta. We're super excited to announce we're a YC company today. We'll be answering questions as much as we can. If you've ever wanted to know anything about drones, technology and agriculture we'll do the best we can share.
Hey Nikhil, I am interested in what you are using drones for in agriuculture. I was using drones about 2 years ago for farming and helping farmers optimize different things and am curious how far it's come since then.
Back then, we were just beginning to see the possibilities. Of course we'd do drainage mapping, NDVI maps (as a yield-correlate), we found success mapping field trials and monitoring field trials.
There were alot of possiblities we could've gotten into, but we found it hard to totally commercialize. For example, trying to map chlorophyll content, or water content. Water content was an easier sell, but I got the feeling that drone-assisted precision agriculture was very much still in its infancy at that point, more hype than substance. I am wondering what you think now? What are you actually using your drones for, commercially, but with specific regards to the actual agriculture decision-making.
Hi James, great question. We have used drones of all shapes and sizes, including off-the-shelf (DJI Phantom, S1000) and custom fixed-wing aircraft (modified X8 flying wing, 11-ft wingspan composite plane).
We're using them to take measurements at the beginning of the growing season (e.g., stand percentages), middle of the growing season (e.g., custom-built sensors to hunt for specific diseases), and finally, tying that back to the quality of the crops that are coming out of the ground. With that last component, we can be very confident when we say our data assists with agricultural decision-making.
I'm curious what the custom-built sensors for specific diseases are? Are you using hyperspectral cameras? I know there was a lot of research into spectral fingerprints of specific diseases.
Also, in regards to the data from the end of season, where you monitor the quality of the crops coming out of the ground. How do you think that assists agricutlrual decision-making? E.g. are you able to predict the quality of the grain and that, in some way, makes it easier to manage?
Hi James. This is Eddie, the CTO of Raptor Maps. We figured out the combination of wavelength bands we need, and combining that with image recognition we can identify particular diseases. We had to do a lot of iterations and correlations to chemical samples to build up confidence that this can actually work. We can only do a few specific conditions today in potato farming, so I can't say this works for every disease everywhere, but we'll keep getting better over time and are working quick to do so. The sensor package is optics-lab style cameras tuned for wavelengths we know have the best signal to noise ratio.
With regards to harvest monitoring, by knowing the exact yields and locations we can correlate remote sensing data to the yield. This type of information will help with management in the coming seasons. But it can also help today. For example, you want to move your lower-quality inventory more quickly so it doesn't affect the good stuff, so you want to put it by the door of the storage facility.
I was wondering when image recognition was going to play a stronger role in the use of drones for agriculture. When I was doing it, it was simply using combinations of spectral bands and not really delving into pattern recognition. Sounds like that's the right way to be headed. Obviously just at the beginnings of it and I appreciate that you know the current limitations of the technology. But still, that's impressive.
Also curious if you are using satellite imagery in any contextual way or to train your image recognition in any way? Or if you are using the lIDAR already available at all? I imagine you are collecting higher-resolution LIDAR when you fly over?
Anyway, good luck with this, sounds like you guys are doing it right.
