Sampling canyons for detecting plants with drone

Hi all,

I am proposing a conservation project for sampling a plant species (Echeveria laui which can grow up to 25 cm in diameter, but I would also need to find seedlings that may be 1 cm in diameter) that is endangered using a drone. I would like to count the number of leaves or the size of the plant every year to observe growth and survival rates. The plant grows in canyons with complex surfaces (different slopes (90, 45, 110 degrees slope), different orientations, etc) and some parts may have low vegetation, so I would think it would be neccesary to take pictures like 2-5 or 10 meters from the canyon wall. Therefore, I have a couple of questions to evaluate if the project would be feasable:

  • Can I use the PIX4D software for planning the flight at such close distances from the canyon wall? I do not know if this kind of software can be able to create a flight plan with such a precise latitude, longitude and elevation data so that I could use the flight plan to re-sample the same area on the next year. If not, I was thinking on manually flying the drone and take pictures. I would also think that I should try to plan the flight to obtain a high overlap between pictures because of the low flying altitude I would use.

  • Could it be possible to obtain images very close from the canyon wall (2-10 meters) and use PIX4Dmapper to make 3D models or orthomosaics to sample plant individuals?

  • Can I use the DJI mini 3 pro drone for doing this task or should I use another DJI drone (mavic or phantom)?

  • If it is possible to make 3D models or orthomosaics as I described above for different years, is there an optimal way to observe differences between plant individuals in the models constructed in two different years to compare plant growth and survival rates between years or should I compare them manually by looking at the the two models generated?

  • Is there a way in the software to know the GPS position or tag in a certain way plants from the 3D model or orthomosaic generated so that I can know in the next year model that I am sampling the same individual and obtain growth and survival per individual?

Any suggestions and comments would be greatly appreciated.


Hi contacto1,

I would not recommend using an automated flight application to control a drone in such proximity to a canyon wall. Non-RTK drones can have many meters of positional error, the GPS signal can be scattered near the canyon wall, and there could be gusty winds when flying that close.

I believe it is possible to manually fly the aircraft between 2-10 meters. However, the key is to have a good overlap between images. You will want to have 70-90% overlap between images. This could be difficult to do manually, but I think if you set the shutter to capture an image every 2-3 seconds, then you can better maintain the overlap as you fly.

The Mini 3 should work for this application. It will capture the GPS location in the images. Then these images can then be loaded into PIX4Dmapper to generate a point cloud, 3D mesh, and/or DSM and orthomosaic that will have georeferencing. You will be able to click on a point in the point cloud and extract the location. Please note that the positions of the outputs will have several meters of potential error. They will be close but not exact.