Summary: We recently did an interesting test where we used NDVI imagery from one of our drones to detect improvements in plant health over a 10 day period.We found that the NDVI was sensitive enough to detect these improvements once the nutrition had been applied.

Main Discussion:

In September 2014 we launched our agricultural drone (www.agridrone.co) at the Agri Mega trade fair in Bredasdorp, South Africa. We are a small startup using PixHawk-based drones to map farms to provide timeous, high resolution RGB and NDVI imagery to farmers, agronomists and soil scientists.

Based on testing work we did the previous season, we were approached by an organic wine farmer and an organic plant nutrition developer to conduct a test of an organic plant nutrition product. The company, Agro-Organics from Somerset West, South Africa, has been working with organic wine farmer Dr Edmund Oettle, of Uplands Organic Estate. They wanted to see if the expected improvement in the vineyard could be detected by our NDVI.

We therefore planned 6 flights over a 10 day period. The plant nutrition was applied at the end of Day 1, after the first mapping flights were done. We then intended to fly on 5 follow up days and compare imagery. In the event, a mistake I made in planning the flights meant that the first day's flight was not usable for comparison, but we were still able to compare the subsequent days.

By flying the identical mission at the same time of day we managed to get comparable images. We were fortunate to get clear conditions on every day, but on some days we had quite strong winds to contend with.

Several images are attached - an RGB mosaic of the vineyard block, showing the 4 rows that were observed. A second image compares the NDVI of the four rows, clearly showing improvement in the treated rows relative to the untreated.

We also took chlorophyll content readings on a sample of the vineyard over the duration of the test, which confirmed the NDVI result.

Usually NDVI is seen as having value in early detection of stress, but I think our project shows that it can add value in identifying improvements too, within limits.

For those seeking more information on our drone, we provide details on our website.

Views: 5453

Attachments:

Replies to This Discussion

wow very impressive! what camera is this? The gopro modified cam?

Hi Mike, thanks for the feedback!

We experimented with various cameras, including the usual Canons, GoPros and keychain cameras. We settled on the Mobius for various reasons - easy to change the filter and refocus, and also the ability to fix the white balance.

Our retail drone has a RGB and a red infra red (RIR) camera, to simplify processing for the user, but these NDVIs are 'true' NDVI that use the pure IR band and the red band, each from different cameras. The registration of the bands is quite labour intensive but in order to pick up these sensitive changes we needed to follow that route.

Hi John,

At what speed and altitude where these taken? How is the Mobius triggered?

Hi Marc

We do table grapes, lucerne, olive trees at 100m+, but the profile of wine vineyards is so narrow that we have to fly lower. This is also a small area so I flew at about 50m at a speed of 6m/s. I'm investigating a new method of mapping wine vineyards which involves aiming the cameras at an angle, rather than from vertical, to capture a larger profile of green leaves and less of the tracks between the vines.

The Mobius does not need to be triggered, it is used in time lapse mode. You just delete the excess photos from the model.

Incidentally, for those who are interested in the results of the plant nutrition, this is the feedback from the farmer himself, Edmund Oettle:

"The harvest is in, it was a good one, and I could definitely see an improvement in both crop size and quality (less disease) on the blocks after treatment. Everyone else I've spoken to has had smaller crops than last year, whereas my crops on both red and white were larger. So I'm happy that I did the application"

Hi John,

Nice results !

Which Software are you using for NDVI calculations ? Which NDVI formuals used ? From which cameras were the images of the attachment taken ? which altitude/speed flown ? In you offer - which camera and software do you supply ?. Thx. Dave

Thanks Dave!

Our system packages PhotoScan, the standard version, which is adequate for generating these orthophotos. We used the original formula and Ned Horning's photomonitoring plugin in ImageJ to generate the NDVI images.

I believe your other questions are answered above - we use a modified Mobius camera with a red filter as the RIR camera and a standard RGB camera for the RGB orthophotos. We flew at 50m and 6m/s for this project but I'd recommend flying higher and faster to cover other crops.

Awesome stuff John! I am doing something similar myself, but without NDVI at the moment as the cameras are so expensive.  I've been considering modifying a Canon S100 myself, but it is still heavier than an action camera. The GoPro's are just too expensive.

I'm busy doing a university project over in the US, and have my own UAV.  Could you share with me, or point me in the right direction, how to modify the Mobius cam? Or if you bought it modified, could you share your retailer?

Baie dankie bru!

Hi Global, we do the modification ourselves but it is tricky and if you get it wrong you can get asymmetric distortion in the image, making lens calibration impossible.

The only place I've heard of selling these cameras modified is publiclab.org. You should check them out.

Hello everyone. I am agronomist engineer, photos assessment NDVI is not my area, though I really want to work on this. There was a very significant improvement in lines 4, however, an improvement is observed in lines 1 (untreated). Will they be by the ends may have occurred some reading error.
I congratulate you for the experience, just so learn to work with this fantastic tool.

Hi Henrique
Given the variation in a range of factors it is impossible to hold conditions exactly constant between days, so we really do not have an entirely fixed control group. However, we selected this image because on balance it shows constancy in the control group versus change in the treated group.
It would be easier to do the test in a greenhouse under hydroponic lamps, where environmental conditions can be fixed, but this was not an option for us.

Thanks John! I'm about to try a DIY modification on a Canon S100, and so I may just give it a bash myself!

RSS

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service