Setting Up a UAV Survey with RS2s?

[intenna (Brian Lindens)]

I had a few questions about setting up a UAV survey with an RS2 and/or M2s and GCPs. Up until now we haven’t needed this level of accuracy so it’s been interesting to figure out how you do it. I will say there may be completely incorrect statements made below if I’ve misunderstood some of the practices. Also, fellow posters and this forum has been immensely helpful. Thanks!

Our setup is two RS2s, a drone, and a multispectral camera (Micasense RedEdge-MX). I’m trying to figure out if using an M2 with a RS2 base is worth it over collecting GCPs. The areas we will be flying are not large (max 150 acres).

For those new to this area like us, I found this thread particularly helpful: Reach RS for GCPs and UAV mapping

In our case we will not have any known points for the base, but we are in Canada so we can submit our RINEX file to NRCAN for post-processing. But we do not care about the global location, just the relative location within the property. So perhaps this processing is not necessary?

My understanding of the UAV survey setup is as follows:

1. Setup up the RS2 Base, allowing it to collect data for 1 hour.

2. Lay out a number of ground control points, in our case I’m thinking 6, around the property. Use the RS2 Rover to collect a point at each GCP.

3. Fly the drone survey using a flight planning app.

4. Post-process the data, using software like Pix4D to create an Orthomosaic. The camera we are using has an in-built GPS which will make processing the images easier.

5. Georeference the images to the GCPs using software like Pix4D or QGIS’ georeferencer function.

At this point, while the survey won’t be centimeter accurate to a global reference system, it will be centimeter accurate relative to the property itself. Given that we will want to come back and use the same base position, our rover and drone will have that level of accuracy going forward. If our clients want to use their own GPS system (likely handhelds like Garmin or tractor-mounted) to investigate specific points, while the survey will not be centimeter accurate for their devices it will be accurate enough they can within a meter of the points (including the the positioning error in their non-RTK GPS systems).

Assuming this is correct, I have a couple of questions:

a) An alternative to GCPs is hooking up an M2 Rover directly to the Micasense camera (instructions here).

Theoretically, we then would not need GCPs as we would have centimeter accuracy just with the RS2 base and M2 rover (relative to the property). But, with this specific camera it seems it would take more post-processing time as the M2 data can’t be written directly to the camera. From Micasense:

NOTE: The Emlid Reach GPS does not output Ublox GPS packets at this time, so GPS coordinates will not be written to the image metadata. A python script is provided below to assist in creating a geolocation file to use with Pix4D, to associate image number with GPS coordinates.

b) Is there a significant amount of post-processing time to get this GPS information? Our specialty isn’t on the GIS side, so I’m trying to create a workflow that keeps GIS processing to a minimum.

More specifically my understanding is that we will be able to upload the images directly into software like Pix4D and because each image has GPS information from the camera it will easily stitch an orthomosaic. Then we would just need to georeference in our 6 GCPs to get a sub-cm accurate image relative to the property. Is this correct or is there significant processing time I’m missing?

[kseniia.suzdaltseva]

Hi Brian,

In our case we will not have any known points for the base, but we are in Canada so we can submit our RINEX file to NRCAN for post-processing. But we do not care about the global location, just the relative location within the property. So perhaps this processing is not necessary?

If you want only relative accuracy, you don’t need to locate the base at a known point and use the NRCAN PPP server for it.

I’m trying to figure out if using an M2 with an RS2 base is worth it over collecting GCPs.

The answer depends on the desired workflow.

Collecting GCPs only makes it hard to predict which amount of points will let you get centimeter precision, so you should be ready to place a lot of them evenly in the site area. If it’s appropriate, you may use only GCPs for your project.

With Reach M2, you can post-process time marks to get precise image centers coordinates and construct a centimeter-accurate orthomosaic. In this case, you would still need some GCPs to check the precision. Usually, 5 of them are used.

a) An alternative to GCPs is hooking up an M2 Rover directly to the Micasense camera. But, with this specific camera it seems it would take more post-processing time as the M2 data can’t be written directly to the camera.

When you use Reach M2 with any camera, it logs photos time marks, not points’ coordinates. The data is logged in the receivers’ memory, and you can post-process them later on to get the precise coordinates.

b) Is there a significant amount of post-processing time to get this GPS information? Our specialty isn’t on the GIS side, so I’m trying to create a workflow that keeps GIS processing to a minimum.

The precise amount of processing time depends on the project. It should not take a lot of time to post-process the results in RTKLIB. We’ve recently published a new guideline on the PPK Mapping with the Reach M2 workflow to make this process even easier for our users.

More specifically my understanding is that we will be able to upload the images directly into software like Pix4D and because each image has GPS information from the camera it will easily stitch an orthomosaic.

For orthomosaic construction, the projection of each photo should be changed from central to orthogonal. The software finds the matching points on photos, calculate their positions and elevations, and constructs orthophoto using this data.

The precise information about image centers’ positions only allows you to have a more reliable orthophoto with fewer GCPs, but the time for these calculations is still needed.