A wildlife biologist is doing a survey where she need to establish a set of images from a bigger area. With the right methods one can make statistical assumptions about the total number of wildlife within the area. Doing it with drones does pose several challenges. In this article I will describe how the software Litchi together with a DJI Mavic Pro drone can be used/tweaked to set up a simple survey.
This particular survey is based on taking pictures along a transect (line). By counting the number of animals in the images, knowing the area covered by each picture and lastly also knowing the full survey area, you have a good starting point for your survey practicalities.
Other surveys could include establishing full strips of images which are then compiled into one big picture. In both cases someone will have to do the counting, or at least supervise an automated counting process.
All surveys have practical challenges. Drones need charged batteries, you will need access to the airspace, fuel for your car, food, helpful assistants, memory cards, storage space on your phone, the right apps and a lot more.
Litchi is a software to set up photo tasks on DJI drones. It allows for detailed control on where the drone flies, which direction it points, its speed, when it takes pictures and a lot more. It does this very well, but for our use we will have to tweak it somewhat.
The first thing you need to do is to establish waypoints for your survey. Note down the coordinates representing the waypoints you will use during your survey. For more complex surveys you could do this using MissionPlanner. Export or note down the coordinates of your survey.
The coordinates for the transects are in this example based on four sets of longitude and latitude values. The figure below shows both coordinates and the lines between them from start to end.
The coordinates and a comma separated file in the litchi format is all you needed. To make things work alright it is important the Litchi CSV-file has all values set up correctly.
The one variable you should keep an eye on is photo_distinterval. When a value is set here it will take pictures with the metric interval until the next waypoint. It starts by taking a picture at the departing waypoint. In the below example it takes a picture every 200 meters until next waypoint. There the distance counting starts anew. Since the distance between the two south eastmost waypoints is 190 meters it only manages to take the picture at the departing waypoint and then resets at the next one when the drone turns north-east again.
The variable altitude(m) decides the altitude of the starting waypoint.
The drone will travel in a straight line between any two waypoints in three dimensions. I waypoint 1 is 50 meters and waypoint 2 is 100 meters the drone will travel upwards along a straight line.
The full list of attributes for this CVS-file is available in the picture below. You can also download it from Github gist here.
Importing the file is straightforward. Open the Litchi mission hub, choose the file and press “Import to new mission”.
The mission will be placed on the map. The file you imported is half of what you need to properly define a project in Litchi.
To make the project available on your phone app you will have to save the project. Choose “Save…” from the missions tab to store the project. This initiates the project settings.
Please note that the projects settings in the mission hub or in your app it indicates a photo capture interval of 100 meters. The true setting for this project sits in the waypoint settings.
Other settings to adjust/check in the project settings.
- Make sure to set “Heading” in the project settings to Auto. This will keep make sure the top of your picture is in the travel direction of the drone.
- Select a cruising speed. Be reminded that the cruising speed can be much lower if you have a strong head wind.
- RTH (Return to Home) a finish action which secures the safe return of your drone.
If you open the waypoints you will find that the setting is according to your imported CVS file. In this case — 200 meters. You will not be able to manually change them. If you unlock the change protection (the red padlock) and open a waypoint the only thing you will achieve by pushing the photo capture interval around is to change it to a value from 100 m or lower.
In hindsight setting capture intervals for waypoints 2 and 4 was not necessary. The distance between waypoint 2 and 3 was less than 200 meters, so only the waypoint itself had a photo taken.
Litchis approach to survey management is excellent. The online user interface allows the survey coordinator to review or set up new surveys, while the field worker/ranger can use the phone app to download the new surveys. This means I can do my work in town while the ranger just opens his Litchi app and loads the survey projects he is meant to work on with the drone.
If you are looking at taking transects with full coverage this could be done as well. You will have to make sure that the photo capture interval is smaller than the image size in the flight direction. You could stitch your images together with relevant software like Pix4Dmapper or Open Drone Map.
The reason for not suggesting using the photo capture interval in seconds is that the drone speed will vary based on wind conditions. A regular interval between the pictures will therefore not be guaranteed.
I also tried using one image per waypoint setting up the waypoints at regular intervals. Litchi in combination with DJI Go tended to overshoot the waypoint having to double back. For each waypoint the drone would always come to a complete halt. The image was shot perfectly at the right spot, but it drains the battery too fast.
I needed to be able to set up a sub sampling survey. In the process I tried 15 other apps including Skycatch, Pix4DCapture, Altizure, MapPilot, Atlas Flight, DroneDeploy, Maply and others. The main reason for their failure to fit the purpose was that they have been built to support a work process where image overlap is required. For making orthophotos this makes sense. For wildlife surveys — not so much. The closest so far is MissionPlanner which has been built specifically to the open source ardupilot platform. It does unfortunately not work with DJI hardware.
What we are seeing here is that the market for apps to build orthophotos, 3d models, elevation maps and other has matured well. I have still not seen anyone tailoring an app for biological surveys be it fauna or flora. In this case I could not wait and found a way forward. It is a bit perky and not optimal.
Should anyone be interested in tailoring a specialised app or survey setup in their existing app I would be more than happy to advice or forward the request to relevant wildlife biologists.
In the meantime I am happy to use the Litchi app. Good work!