In this deep dive conversation, we unpack what it is that GeoWing Academy is all about! What the mission is, what its all about and how easy it is for anyone and everyone to upskill themselves in all things drones and GIS.
The cape dune mole rat is an interesting creature. Mole rats usually live in social colognes but not the large Cape Dune. These fellas are viciously territorial, liver on their own and love keeping to themeselves.
By looking at cape dune mole rat behaviour observed by Spinks, A. C., Bennett, N. C., Jarvis, J. U. M. (1999). Home range size and habitat use in the Cape dune mole-rat Bathyergus suillus. Journal of Zoology, 248(1), 39–47. and Bennett, N. C., & Faulkes, C. G. (2000). African Mole-Rats: Ecology and Eusociality. Cambridge University Press. GeoWing Academy designed a method of using drone photogrammetry outputs and GIS software to track burrowing activity over 1 week in 2024 and 1 week in 2025. The results are interesting to say the least!
See the latest vlog below for more details!
Alien Invasive Plants (AIP) have become a major threat to South Africa’s sensitive Fynbos biome. In 2017 and 2018, fires in the Western Cape region killed 8 people and destroyed over 2000 homes and devastated biodiversity in the region. The intensity of these fires was amplified by the massive amounts of AIP, in particular Black Wattle (Acacia mearnsii), Gum (Eucalyptus sp.) and Pine (Pinus sp.) that have gone unchecked and uncontrolled for decades. Now in 2025, the problem is even more pronounced and current methods of monitoring and clearing AIP are very inefficient, very time consuming and very costly.
Drone technology and open source software can be used to map, locate, classify age, determine ease of access, determine urgency for clearing, define burn intensity and then plan removal projects based on this information. Drones can also be used to clear AIP by using precision spray methods to kill off very dense stands. This coupled with ground based removals will drastically improve current methods and we may actually have a small chance of regaining the biodiversity lost.
This presentation was hosted by the Save Wild Project. For more information have a look at the presentation given to Western Cape municipality, government and communities on the subject of tech based applications for UAVs and AIP control here:
ere Machine Learning and RGB drone data for plant health? GeoWing Academy decided to take a look at how effective using a custom built machine learning pipeline and RGB derived plant health indices for crop health analysis. The results are SUPER interesting!
A strawberry patch became the unwilling test subject for this experiment and two data sets where captured; one set on a cloudy day and one set on a sunny day. The machine algorithm was trained to find each strawberry within the patch and then compute each individual plants Modified Photochemical Reflectance Index (MPRI) value.
The MPR Index is an RGB-based vegetation index that compares green and red reflectance to estimate plant health. It produces higher values where plants reflect more green and absorb more red — a common sign of healthy, photosynthetically active vegetation. Higher MPRI values indicate more green reflectance relative to red, which often corresponds to healthier vegetation (since healthy plants reflect more green and absorb more red due to chlorophyll activity). Lower MPRI values suggest more red reflectance relative to green, which can indicate stressed or sparse vegetation, or non-vegetated surfaces.
The overcast scan MPRI statistic mean of 0.087 indicates that the plants in the patch were moderately healthy which ties in well with the ground truth evaluation done by the farmer. The patch was then subject to cleaning (all dead and dying strawberry leaves removed) between the two scan dates, the next scan being conducted on a sunny day. The sunny day scan bore some very interesting result indeed!!
This index will not be as rigid as multispectral imagery and will only display surface (visual) health differences but it certainly has its place for crop management where time and money are a factor of production. So can it be a useful tool? Find out in GeoWing Academies latest vlog:
15 years ago I was managing a game reserve in the Mapungubwe area in North Limpopo South Africa. The reserve had not had a permanent manager and there was some serious need for maintenance on fences, roads and dams. In fact there was only one of the four dams on the entire 3500ha reserve that had water in it. The rest had been damaged but flooding and neglect. One of my first tasks was to re-establish these vital water points. In order to do this, I had to use what’s known as a “Dumpy” level, a piece of equipment that allows you to read terrain levels in order to make sure that whatever you are building comes out level and straight, in this case it would have been the dam walls. Now you must understand that I had never used this sort of fancy pants equipment, so the learning curve was steep and fast. I was used to using a piece of string and a spirit level, so this new tech was a bit challenging to get right initially and walking back and forth across broken earth dam debris in 35 degree C heat to take measurements was time consuming. Needless to say, more colourful phrases other than “dam it” were thrown around as I fell over a number of times traversing the treacherous terrain with the cumbersome equipment. I remember thinking back then that I hoped the future would have easier ways of doing this sort of work.
Image 1: One of the many broken dam walls on the property. This particular one, although not a true “earth” dam was destroyed during a flood. The erosion damage caused by the breach also had to be filled in and brush packing used to prevent any further erosion should it have rained again.
It would have been nice to be able to calculate the amount of material needed for the construction of the new dam wall for budgeting purposes but at the time I had no real survey knowledge to be able to work this out properly, so I gave a rough estimate and got on with it. I don’t like doing this because budgets are usually very tight when running game reserves. Over spend here, loose out there and with so much that needed to be done, there wasn’t much wriggle room at all.
Fast-forward to 2023 in another part of the country. I was asked by a member of the local farming community whether or not drones can be used for surveying potential dam sites that would be used to irrigate lucerne fields and orchards. The intended areas were shallow valleys that, should the earth dam walls be built, would flood providing hundreds of thousands of cubic metres of precious water to the surrounding farms. Well, that was the thinking anyway but the farmers could not be sure if these figures were true just by eyeballing it and wanted surveys done of each site to see which would yield the correct amount but they had a very limited time frame in which to do it. Having done dumpy level surveys on dams much smaller than the intended dams of this project, I knew that traditional survey methods would not work within their limited timeframe at all as they would take many days for each site. The terrain was rough and at least two trained surveyors would be needed to survey each site (surveyors for these sorts of projects come with a very high price tag as well).
Image 2: This drone image is from another site but shows what a dam would look like once completed and shows the dam wall length, (108.5m) the flood plain perimetre length (excluding dam wall, 547m) and the volume of the dam (92 972 cubic metres).
Challenge accepted! Admittedly, at this point I had not done a potential dam site survey with photogrammetry outputs but, as we know, when you are working with powerful GIS software and centimetre GSD (ground sampling distance) 3D data, you can do almost anything with regards to geospatial information acquisition. It took nearly five batteries for the DJI Phantom 4 Pro V2 (roughly 80 minutes of flying time) to collect the required data per site, averaging 1000 images per site, covering 59ha and 70 ha respectively. This turned out to be overkill but it is always better to collect more data than to find out you are short and have to redo the flights. It took about 3 hours of processing time per site to render the Digital Terrain Models (DTM) and high resolution orthomaps needed for the GIS site analysis. High precision georeferenced accuracy was not required for this project as the only numbers that were required where the flood plain areas, potential dam volume and potential dam wall length. (See https://geowingacademy.com/did-you-know-about-elevation-and-gps-offsets-between-drone-data-collected-on-different-dates-and-how-it-effects-photogrammetry-processing/ for more information on Relative vs Absolute drone map accuracy). Essentially, building a digital “virtual” dam was the name of the game for this project.
Image 3: Full scan area (70 ha) with p roposed dam outlined in yellow (926 064 cubic metres)
The GIS analysis part was a fairly straight forward process (well at this point, I had done cut/fill surveys for mines prior to this project so some of the GIS skills needed had already been learned prior to this project). The farmers had stipulated that they wanted a 50cm contour map as well as the volume, area and perimeter information. So 50 cm contours were computed using the DTM. Two orange road cones were placed where placed on the ground on either side of the valley on each site to show where the wall would be built between before the drone flight. These were then used to draw a digital “wall” and the contour line that intersected the potential dam wall was used to calculate the floodplain area and dam perimeter length. The potential dam volume was then calculated.
Table 1: The dam parameters calculated in GIS.
So everyone was happy and after 2 days the reports were sent off and the job was done. But I wasn’t satisfied. There must be a way to calculate the façade area of a potential earth dam wall as well which would allow for the calculation of materials needed for which can be factored into the budgeting process. So, with some assistance from an engineer, who asked me if it was possible to work out a number of different things in GIS in terms of horizontal and vertical measurements. I had not done this before but I knew it was possible (anything is possible with this amazing technology) so it was back to the drawing board and I set about how to work out a method for acquiring the various measurements that were needed. He then took these measurements and showed me how these can be used to calculate the façade area of the dam wall. This was brilliant! When you combine this information with the known measurement requirements for stable earth dam wall construction, you can calculate the estimated amount of material needed to complete the job quite accurately, so when it comes to compiling the budget report for the dam projects, Bobs your Auntie! You can learn how to quantify dam measurements with GeoWing Academy’s DAM IT! Course, a two part deep dive into all things dams!
When looking at orchards and plantations, the short answer to this question is “Much less than you think”! The long answer: “How to go about it?” Let’s dissect that a little shall we?
Data collection using an off-the-shelf consumer drone that is capable of functioning with either proprietary or third party mission planning apps is a fairly simple process:
Let’s say we want to count the number of trees in our newly planted orchard, inspect their canopy health, and acquire each plant’s GPS coordinates. The field is fairly homogenous, there are a few grasses, ferns, weeds and sedges between the newly planted trees, but for the most part the rows are clear (it’s important to consider when to fly and that the cleaner your target area, the cleaner your computations). The tree canopies are fairly well established, so a GSD of ~3cm per pixel will suffice (90-100m above ground level flight level when using a DJI Phantom 4 Pro V2).
Processing the data is a fairly straight forward process too; in WebODM simply using the default “High Resolution” settings will generate the RGB orthomaps, Digital Surface Model (DSM) and 3D point clouds required for later use in Cloud Compare and in QGIS along with the machine learning algorithms from the Orfeo Toolbox.
Goody! So we know what our project, data capture and processing requirements are and we have our maps and models ready.
Figure 1: Scan area, roughly 3ha. Mostly bare ground, good tree separation. Weeds, grass, sedges and ferns have started to encroach in some areas.
Now we need to download the Orfeo Toolbox libraries and get our hands dirty for a minute. These powerful machine learning algorithms allow us to customise our output requirements and make fine adjustments to the machine learning process which will help when it comes to isolating objects within our RGB drone data sets. These algorithms work best when the target area has well contrasting vegetation and definitive objects as it is easier for the algorithm to classify pixels. So for orchard and plantation detections, these algorithms are usually able to detect the target trees with a high level of confidence and accuracy. You just have to know how to tune them correctly.
Adding layers of depth for the algorithm to read off also improves the detections by giving the algorithms more information. This can be done by adding vegetation health index maps and tree canopy height models (CHMs) extracted from 3D point data. When working with plants and when you have access to multispectral drone data (Near infrared and red edge) you can add these layers or create index models using the multispectral information to create a good level of depth for the algorithm to feed on. In this example, only RGB data and RGB indices were used.
Figure 2: Overhead view of the canopy height model (CHM) generated using 3D point clouds (scale bar on the right in metres)
Figure 3: Side view of the CHM showing terrain gradient as well
Figure 4: Plant health index
Now we need to train the algorithms what to look for; this is effectively a “supervised classification” where you are telling the algorithm what-is-what in the image, then isolating the required object from the “noise”.
Figure 5: Training polygons that tell the algorithms what to look for in the image
Once the tuning has been tested and you are satisfied with the detection accuracy and results, you can then decide on the data you would like to output from the detections, such as plant health for each tree, canopy area, tree spacing, GPS location, average tree heights etc. The big advantage of open source software is that it is highly customisable and it is very easy to design a model that spits out the specific data you need.
The importance of layering and choosing and tuning the correct algorithm for the job when you compare detection data becomes clear.
Figure 7 (right): Basic detections, no layering using the default settings in orange
Figure 6 (left): Target area with mixed vegetation. The similarities in vegetation RGB colour of some plants may cause confusion for the algorithm (a fern might be misidentified as a target tree for example). This is why layering, tuning and using the correct algorithm with the right amount of training information is important.
Figure 8 (left): Basic detections, no layering some tuning in blue over orange (note the slight difference of the detection, excess incorrect pixels are now excluded).
Figure 9 (right): Finer tuning green over orange and blue.
Figure 10 (left): Finer tuning, additional training data and the introduction of an additional vegetation health index purple over orange blue and green.
Figure 11: Finer tuning, additional training data and the introduction of an additional vegetation health index using an alternative algorithm that was finely tuned (in red). Note how the algorithm is now eliminating all non-target tree pixels and only finding the correct object.
Once tuned, the pipeline is ready for repeatable use with very minimal input required. The drone maps go in, you tell it what to look for, hit run and the data tables come out the other side. Super easy! The tuning may take a bit of time initially, but once tuned you don’t have to tune again! So lets say you want to look for alien vegetation in indigenous forest during the flowering season, find and count sea cucumbers on a shallow reef or look for ground bird nests in open areas, the process becomes a lot faster and simpler than having to do so with traditional methods only (it is important to note that you must still ground truth the outputs and don’t rely soley on the technology! Drones and GIS are “force multipliers” and not “silver bullets”).
Figure 12 (above): Tree detections marked by points after the detection pipeline was tuned and automated. The process needs to be smooth with as minimal input and as repeatable as possible. So: data maps in – train model – data tables out.
Figure 13 (above): The confusion matrix shows how confident the algorithm was with detecting trees. It is not possible to get 100% with RGB data or even hyperspectral, so this needs to be factored into the statistical analysis. More complex maps with very complex scenery and close colour matching between species means confidence may go down.
Figure 14 (above): An example of the data tables outputs that are generated through the custom pipeline.
So what does all of this have to do with cost? Well let’s break it down. An off-the-shelf drone may cost a few hundred dollars and a mission planning app that automates the drone’s flight path for simple mapping missions is either free or has a small monthly fee of a few dollars. So that is the hardware side of things. Then there is the photogrammetry software; with WebODM, the installation driver is extremely affordable compared to proprietary software such as Pix4D or cloud-based platforms such as Drone Deploy, which can easily go into the thousands of dollars for the subscriptions (don’t get me wrong, these are phenomenal platforms but are not necessarily in the range of a farmer’s or a conservationist’s budget). It is also free if you have the tech skills to install the drivers yourself. Now on to the GIS software: well QGIS is totally free (donations to the project are a must if you are to use this software though, without these brilliant minds giving their time to building this software, we wouldn’t have these tools so please donate) and so is the Orfeo Toolbox with all the machine learning libraries. The cost of training is also inexpensive and within a few days you can go from zero understanding of drones and GIS to creating your own custom machine learning pipelines for your project specific needs (go to www.geowingacademy.com for more information).
So for under USD$1000 you can capture, store, analyse data, build and customise your own reusable machine learning process that can be used over and over again with minimal inputs for massive outputs as well as keep digital records for later comparative use.
Lastly we have to factor in time (after all time is money). How much time and therefore money do you save by using technology for data capture and analysis compared to using traditional survey methods? The answer to that is…priceless!
