Hi all
I am learning the ways of TerraSolid. In this software I have licenses for TerraScan and TerraMatch. I am trying to improve the vertical accuracy of my first LIDAR mission.


So far I found this helpfull video showing me how to do so using GCPs: Processing DJI L1 Point Cloud with TerraScan UAV and TerraModeler UAV (from 15:30 mins).
I have 4 usefull GCPs in the area, each measured using a mobile RTK GPS. When I follow the steps from the video (using the Output Control Report function) I receive the following results afte9M4bss8PETn4LrDqn0+B9D4lqKT3I7AOistOeucNlfRUvAvk/Pi5Se9jgEp6Hwa2SW+N/w/WSf+DaMI6cQAAAABJRU5ErkJggg==[/img]


As you can see, there is still an average difference between of 15 cm between the real GCP height and the point cloud height. If we take a deeper look at a GCP we get the following (Picture 2):





- The little rectangle contains the real altitude of the GCP in the 3D space.
- The group of points above are the measures GCP points captured in the image (called laser points in TerraSolid)
- The blue line is the corrected height of the GCP following the above strategy.
- Don't look at the giant X, that is my cursor.


What we can see here is that there indeed a mismatch between the three features. My question now is twofold.
1) What other actions can I perform to improve my z-allignment? I tried looking in the manual (https://terrasolid.com/guides/tscan/introcoordinatetransformations.html?anchor=sec and https://terrasolid.com/guides/tscan/prjoutputcontrolreport.html) but I don't see any other methods using my 4 GCP points.


2) If my z-alignment were to be improved, what exatcly am I looking for?
      a) Do I want the blue line to be in the GCP point cloud (= my laser data)?
      b) Do I want the blue line to be at the same level as the rectangular pixel (aka the real GCP evelation)?
      c) Do I want both the blue line and the rectangular pixel to be in my laser data?


Is there anyone having experience in this program who knows the answers to these questions? Only if I have these I will be able to construct a DEM knowing what it is actually showing.


Many thanks for your replies!

"The blue line is the corrected height of the GCP following the above strategy"
When I see this kind of thing, I think that the problem does not necessarily come from the software.
A GCP is not intended to be corrected. It must be used to correct other data. That's the only thing you can do with it.

First you need to calibrate your instruments. Calibrating does not necessarily mean that there is a complicated process to apply via software, it can be done only with the help of a comparative measurement.


To do this, there is only one reliable method: to measure a known geodesic reference point.
GDOC
You take your "RTK GPS mobile", whatever it is, and you will measure the point that you have chosen and there you will be able to say what is the margin of error of your GNSS receiver (rover) and base station pair. At this point you will be able to confirm the dimension of the pole and the phase center of the receiver antenna.

Then you can do the same thing with LiDAR, by surveying a geodetic marker.

With this, you are already starting on a solid basis, which will prevent you from having to go into GCP corrections.

Then, we can't see very well what is written on the screenshots. I suggest you edit your post and re-upload them in large format.

But from what I see, it's very surprising that you get such a level of different error with a LiDAR.
I mean, having 15cm difference everywhere, uniformly across the entire cloud point is ok, because that's what you'd expect with a LiDAR.
But having 15cm on one side and 5 on the other is something common in photogrammetry, but not at all in LiDAR.

Even if here we are talking about L1 from DJI, which is not the best. Reflectivity of your GCP may have an impact.

Afterwards, yes, there are functions in the software to make the point cloud stick to the GCP or vice versa, but this will "bend" the data in one way or another, over which you have no control, and which will not necessarily be qualitative.
In TerraSolid i don't remember, but in Global Mapper its something like QC control something like that.

Just in case, didn't you check the "smooth point cloud" death box in Terra?

LV_Forestry Posted at 12-19 10:07
"The blue line is the corrected height of the GCP following the above strategy"
When I see this kind of thing, I think that the problem does not necessarily come from the software.
A GCP is not intended to be corrected. It must be used to correct other data. That's the only thing you can do with it.

Thank you for your rapid reply!


I did not check that box in DJI Terra. I am using the L2 sensor and uploaded a better version of the photo.
The suggestion of capturing know geodesic reference points is very helpful. But how do I perform this in practise? If I am working in a salt marsh and the nearest geodesic reference point is 1 km away, do I fly my drone all the way there and take 1 lidar strip? With my RTK GPS (sorry I forgot the name, but will ask a collegue) I can easily do the same. As for the data it recorded, you can see it has a z-accuracy of 1.2 - 2.7 cm.

Point number	x	y		z		Fix type	Satellites used	Deviation x	Deviation y	Deviation z
1	132978,1	226793,2		4,977		RTK	37	0,005	0,006	0,012
2	133013,8	226840,9		4,731		RTK	34	0,008	0,011	0,019
3	133018,6	226781,8		4,693		RTK	35	0,011	0,015	0,025
4	133041,2	226732,2		4,861		RTK	36	0,012	0,016	0,027

It is indeed very strange that some of the real GCP locations lie close to the point cloud location, sharing the same dz (GCP 1 & 2), but other (nr 4 & 5) lie far away and have an opposite direction of elevation difference (one is overestimated by 15 cm, and the other underestimated by 30 cm).

Am I missing something here?

WETCOAST_JC Posted at 12-20 00:42
Thank you for your rapid reply!

Well it's not that bad. There is clearly a problem with GCP 4. It's hard to tell you what.I would start by taking out the GCP4 from the project and work only on the remaining three. Then you take the RMS given by Terrasolid and that's it. After that you can publish your raster and go double check physicaly if the values ​​are correct.

But indeed doing the survey of a geodetic marker is not a luxury. If the problem is hardware, GNSS receiver or IMU, this will allow to highlight it.

The difference in altitude can come from the WGS to geoid conversion. Especially in the case where you would have used a grid file that can be interpolated differently. Which will give minimal errors of a few centimeters. But not 30!