Precision Agriculture for Development (PAD), in partnership with Levitation Dynamics, set out to conduct a feasibility study on the usefulness of drones in remote sensing for agriculture.
With PAD having already begun a few SMS-based advisory services, as well as having some reliable contacts in the Eastern region, it proved to be a good area to conduct this pilot project. The initial plan was to cover numerous sites in different towns, but we later rescheduled.
The main objective of this pilot mission was for PAD to test the viability and potential use cases of UAV remote sensing in precision agriculture and smallholder farming. PAD had earlier last year made recommendations to farmers for different types of fertilizer to use, and had wanted to explore the possibility of collecting quantitative data on the same. It also served as an opportunity for us, Levitation Dynamics, to test our capacity to deliver value in agriculture using drones. The crop to be analysed during this mission was maize, a crop that responds very well to drone mapping, as is the case with most cereals.
Our first site visit took us to a rural area a few kilometres off the main road leading to Machakos town. We crossed a few ridges and hills to get to the location, set up our equipment, planned our flight path, and took flight, much to the amazement of the locals, as was expected.
The drone’s primary imaging sensor, a Normalized Difference Vegetation Index (NDVI) based camera, captures light in the Red, Green and Near Infrared spectra (RGN), and is set to take pictures of the landscape at regular intervals, and the flight path is configured to ensure adequate overlap between the images captured, to ensure accurate orthomosaic stitching for post-processing.
Once this was completed we set off to our next site in Mwala, a 45 minute drive from our first location. Here, the farm was moderately larger, and as such required more time for the images to be captured.
After a quick lunch break, and a few more brief trips to some other sites, we made our way back to Nairobi, where the data processing would be done. Photogrammetry and orthomosaic generation are very resource-intensive operations, requiring specialized software and hardware resources, well in excess of what is normally available on consumer equipment.
The workflow involves first processing and calibrating all RAW images into georeferenced TIFF files, as this format is able to retain the essential reflectance data used to calculate NDVI values and gauge crop health. These images are then aligned and stitched together to form dense point clouds, 3D meshes, Digital Elevation Models (DEM) and Orthomosaics, all of which are used to derive valuable data and insight, as further shown below.
The orthomosaics, in particular, are quite handy pieces of information, as these actually contain the data used to calculate the NDVI values. To briefly explain NDVI; the sun’s rays contain both helpful and harmful radiation for plants. The green substance on plant leaves, chlorophyll, is able to absorb the helpful (light) radiation, and reflect away the harmful (infrared/heat) radiation. A measure of a plant’s general health is in it’s ability to reflect this harmful ratiation, information that is captured perfectly using specialized imaging sensors. Healthy plants show up with a high index value (green color) while drying plants and bare soil have a much lower index value, and show up with a more orange-red color. This allows a farmer or agronomist to identify potential problem areas such as inadequate water or fertilizer on a farm; information that may not be immediately apparent at ground level, especially at a larger scale.
In the image below, it can be observed that the farms closest to the homestead (Farm 1) show up as having healthy plant life, whereas the crops to the far left of the farm (Farm 2), across the ridge, seem a lot more neglected, and as such have much less vibrant plant life. This allows farmers to know where to focus farming resources, and is increasingly more effective with agricultural projects covering a wider area.
The data collected at the second site was similar in nature to the first, showing detailed topography and reflectance data useful for our analyses.
We were happy with the vital data we were able to collect through this pilot project. The raw data we captured can be processed and formatted in a multitude of ways, and so we will continue to experiment with it and see what kind of information we can come up with.
After further discussion with the PAD team, we found it would be important to look into an even wider range of value-additive services. These include accurate yield estimation, distinct differentiation between the various threats to plant health, among others, and these would provide even more quality insight for optimization of farming practices, particularly for the smallholder farmers.
Generally, this UAV remote sensing pilot project can be deemed a success, and we are grateful to Precision Agriculture for Development (PAD) for organizing the logistics and resources, and their tireless efforts in improving the agricultural landscape in Kenya and around the world. We would also like to thank the farmers for welcoming us into their spaces to conduct valuable research that stands to benefit millions of others. We are very excited to see what is in store for the drone industry in Kenya and the region, and we are always happy to contribute towards seeing the industry grow from strength to strength.