NDVI for assessing crop health in agriculture using Drones

Drones for NDVI

NDVI has been around since 1973 (Rouse et al), if you're reading this blog post I have to assume a wealth of farming and agronomy knowledge on your part. This section is to fill you in on the UAV developments. How have drones changed the way we use NDVI and what can you expect from Sentry?


Landsat has been delivering free satellite imagery every 16 days to the world, available online. Why don't farmers use this? Take a look at the example image to see why. Landsat is great for checking the biomass in the rainforests and getting a macro feel for indices of greeness over the planet as a whole- but for diagnosing problem areas in a few thousand acres of farmland it's not the answer.

NDVI from landsat- enhanced
NDVI from landsat- enhanced

Here is a 20 acre field from the Landsat NDVI imagery- the resolution is not good enough to be useful.

Sentry UAV Drone captured Field example
Sentry UAV Drone captured Field example

The resolution is good enough to make inferences about crop health, seed density and normal, RGB images are supplied alongside for comparison in even higher resolution.

UAVs, RPAs or Drones will be fitted with cameras holding sensors able to take photos in the 8 to 20 megapixel range- depending on the light spectrum they are capturing. Why?  When capturing the darker NIR bands, the same size sensor is used but with fewer, larger pixels on board so that they can capture more light in a shorter space of time- because the UAV will be moving at around 25 knots. This keeps the image quality crisp and high quality. The resolution of the aerial photography is dependant on the altitude that the drone flies.  The lower the better image quality, however each photograph will be of a smaller area of ground and so more photographs will be taken- and the mission will be of a greater duration.  So there is a trade off between the quality required and the time the aerial platform will be able to fly.  Sentry Multirotor Drones can capture individual blades of grass at 50m AGL (above ground level) and cover around 20 acres per flight at this level of detail- which is generally over the top but impressive. Less detail and more coverage is the normal order of business- around 3cm per pixel. Fixed wing drones can fly for hours at any legal altitude and are more suited to the task- though they require a far higher level of operator skill so you will rarely see them in use in the UK. We use both flight platforms.

How high resolution is a drone survey?
How high resolution is a drone survey?


This image is compressed heavily for the web, the raw survey data is extremely large!

The Index

(It hasn't changed)

NDVI (Normalised Difference Vegetation Index) is calculated as: (NIR-Red)/(NIR+Red)

This formula is applied to an orthorectified stitched NIR-R image of the subject, the result is a black and white image with grey values between 0 and 255- notionally between 0 and 1. There are countless off the shelf software packages that make the calculations automatically and apply the false colourations automatically. My current opinion is that these are not yet fully developed and have been added to cloud based mapping packages to improve competitive edges.

I prefer to use GIS software and a raster calculator, and then apply the red, yellow and green channel values in a bespoke manner to ensure that there is adequate contrast separation between the channels. The automated software packages tend to lose precious resolution in the conversion. As software gets patched daily- I'll keep on top of developments and ensure Sentry has the best solution to make sure customers are getting the clearest results. Currently, manual oversight provides the best outcome.

Comments are closed.