Category: Agriculture

24 Sep 2018
Wheat multispec

Multispectral Data for Agriculture 2018

This year we have been working with a few local farms creating agricultural maps using drones to assess plant health and assist variable rate application of fertiliser, herbicide and pesticides. Using a modified Phantom 3 drone with the Parrot Sequoia multispectral camera we successfully performed dozens of missions this summer on a variety of crops. Some examples are shown below processed using Pix4D and the ATLAS platform from micasense.

Slug damage and another issue
Top: Wheat slug damage over winter. Bottom: Potatoes with some compaction lines and density variation.
Variation in results for NDVI and NDRE for potatoes.
Variation in results for NDVI and NDRE for potatoes.
Variation between NDVI and NDRE for wheat.
Variation between NDVI and NDRE for wheat.

One of the biggest problems we faced was presenting this data in a reliable GIS software. Most of the solutions currently lack features and only do one thing very well. For example ATLAS from Micasense is a far superior visualisation tool than Pix4D but it cannot create variable application zones for tractor shape files.

Towards the end of the season we purchased a Sentera 4K Double Multispec camera which has some interesting features over the current multispectral cameras on the market. In particular it has a higher resolution camera, is very compact with only two lenses and comes with a DJI compatible gimbal. This means more reliable stitching as the camera is always NADIR (looking straight down) where the P3 or fixed wings will capture slightly angled images depending on wind.

Over satellite imagery aerial imagery from drones provide far greater resolution. For example ESA Sentinel-2 has ground resolution down to 10m only and updates every 10 days which might or not be cloud free. A drone at 100m with the Parrot Sequoia camera provides GSD of 10cm and for the Sentera 4K 3.5cm, regardless of cloud cover.
There is certainly room for both approaches in farming, for analytical strength aerial imagery outperforms satellite. We look forward to sharing more imagery captured soon.

09 May 2017
Phantom 3 AG

Agricultural Monitoring Trials

We’ve been refining our agricultural monitoring service which we will offer to farmers in Scotland and Northern England in the next few weeks. We have combined the MicaSense Sequoia Multispectral camera with a modified UAV for data collection. Using the Pix4D Mapper Pro software we can now produce high quality field data to supplement modern precision agriculture.

Currently there are several NDVI survey cameras on the market but many of these lack features and require difficult post processing to calibrate and geotag the resulting images. The Sequoia camera provides an all in one solution and integrates seamlessly with Pix4D Mapper Pro.

Phantom 3 AG
Skytech Aerial Phantom 3 AG In Flight.

This light weight camera made it possible for us to modify a DJI Phantom 3 drone to carry it. Over larger UAV platforms the smaller Phantom class is light and efficient to fly as well as easily transportable when walking a long way over rough ground. We’ve flown hundreds of acres so far in our trials and it’s proven capable of capturing the imagery required for agricultural mapping.

Once we have acquired imagery data from the field it is processed using Pix4D Mapper Pro a professional mapping software that calibrates, combines and indexes thousands of images to create an indexed field map. By combining different light spectral bands red edge (RE) and near infra red (NIR).

Vegetation Reflectance Bands
Vegetation Reflectance Bands

The principle of NDVI analysis is that a healthy plant reflects more NIR than an unhealthy plant. Plants may look green but the RE and NIR bands of interest are unseen to the human eye. These along with RGB colour values may be added to different indices formulas to provide a plant health indicator.

Pix4D AG
Screenshot of Pix4D Mapper Pro showing the stressed areas in red.

The indexed map colour then gives a plant ‘health’ indicator from green for healthy plants, through to red for stressed plants. This data can be used to create an ‘agricultural prescription’ which is simply an application map file readable by modern farm equipment so that they can spray the correct amount of fertiliser or pesticide in the marked areas.

This health report can then be used along with farm based data (what, when has happened already) as well as point inspections to diagnose the problem (pest, nutrition, irrigation).

Further examples follow:

Example 1
Example 1
Example 2
Example 2