So from what I noticed so far, because I cannot export the POS data and inport it back with corrected coordinates, I cannot achieve the same precision as with aerial.Also when using L1, there is no option to use GCP, so how do we achieve cm precision. I guess we need to always use the DRTK2 but then again that's a huge headache to set up every time.
My only solution is to have the RTK with fotogrammetry and align the point cloud in a separate software.

check out the attachment. The model with orange is the correct one with RTK. The grey one, is the L1.  

You can use NTRIP for RTK w/o needing the D-RTK 2, provided you have internet connection and there is a base nearby providing corrections.

How many meters is the difference?
It's a safe bet that in DJI Terra the vertical datum does not work for L1. I also have this problem. The datum exists in the list, it is selectable, but does not lead to anything because the geoid does not seem to be referenced. I make the correction in post-process in QGIS and it is no longer a problem. If your GCPs are in ellipsoidal height, it becomes more worrying indeed.

patiam Posted at 9-1 07:22
You can use NTRIP for RTK w/o needing the D-RTK, provided you have internet connection and there is a base nearby providing corrections.

I totally agree with Patiam. But I think that's what he did, otherwise it's not possible to extract the point cloud via Terra. You necessarily need a link to an RTK base station throughout the flight, or a PPK observation file, also covering the entire flight.

LV_Forestry Posted at 9-1 07:25
How many meters is the difference?
It's a safe bet that in DJI Terra the vertical datum does not work for L1. I also have this problem. The datum exists in the list, it is selectable, but does not lead to anything because the geoid does not seem to be referenced. I make the correction in post-process in QGIS and it is no longer a problem. If your GCPs are in ellipsoidal height, it becomes more worrying indeed.

Ahh, good cal LV_Forestry. I don't use Terra so can't comment on it's geodesy capabilities/limitations. But that sounds like it may be a known issue...

LV_Forestry Posted at 9-1 07:25
How many meters is the difference?
It's a safe bet that in DJI Terra the vertical datum does not work for L1. I also have this problem. The datum exists in the list, it is selectable, but does not lead to anything because the geoid does not seem to be referenced. I make the correction in post-process in QGIS and it is no longer a problem. If your GCPs are in ellipsoidal height, it becomes more worrying indeed.

around 30 m. All the flight is in fixed RTK, even using the photos from the L1 I can have good accuracy, but not with the lidar point cloud.

FlorianGhe Posted at 9-1 21:05
around 30 m. All the flight is in fixed RTK, even using the photos from the L1 I can have good accuracy, but not with the lidar point cloud.

If you are in the East part of Romania, 30m is a geoid problem. If not, then need to search deeper.

LV_Forestry Posted at 9-1 21:27
If you are in the East part of Romania, 30m is a geoid problem. If not, then need to search deeper.

yes I'm in eastern Romania, but what geoid should I choose?

patiam Posted at 9-1 10:35
Ahh, good cal LV_Forestry. I don't use Terra so can't comment on it's geodesy capabilities/limitations. But that sounds like it may be a known issue...

You are very lucky not to have to use Terra. With the L1, you have no choice, you have to go through Terra to obtain a .LAS point cloud. On paper and in the commercial promise, it's fantastic. In practice the name of the geoid exists in the list, it is selectable, the report even confirms that it is the geoid used... bull.... Maybe it's me who's too dumb, but I don't have time to take care of implementing a geoid in Terra so I'm doing the correction on QGIS in post process.

FlorianGhe Posted at 9-1 21:28
yes I'm in eastern Romania, but what geoid should I choose?

Romania MN75 height seems to be the lastest available. Unfortunatly it is not in Terra list. Doesn't matter, just think about applying the offset in post process.

what about 3d model, did any of you managed to export 3D model with coordinates?

FlorianGhe Posted at 9-1 23:04
what about 3d model, did any of you managed to export 3D model with coordinates?

Mydocument/DJI/DJITerra/user/project/las your point cloud is there.
Then CloudCompare or GlobalMapper, Filter the point cloud, build a mesh. 3D model is ready

LV_Forestry Posted at 9-1 23:30
Mydocument/DJI/DJITerra/user/project/las your point cloud is there.
Then CloudCompare or GlobalMapper, Filter the point cloud, build a mesh. 3D model is ready

Thank you. I already have a similar workflow. It's a shame we cannot directly use the mesh generated with terra. For example in Agisoft Metashape I can export it without issues, while Terra does not have coordinates in mesh but only in point cloud.
I already written to DJI but they replied that they will look it up.

FlorianGhe Posted at 9-1 23:37
Thank you. I already have a similar workflow. It's a shame we cannot directly use the mesh generated with terra. For example in Agisoft Metashape I can export it without issues, while Terra does not have coordinates in mesh but only in point cloud.
I already written to DJI but they replied that they will look it up.

DJI's standard answer.

LV_Forestry Posted at 9-2 06:06
DJI's standard answer.

Just got this one:
"Only the elevation correction model downloaded from https://cdn.proj.org/ is supported to be imported into the vertical coordinate system of the existing list name. Custom elevation correction models are not supported currently."

The flight control for the L1 seems to optimize for the accuracy of the LiDAR data (as it should) but that leaves the RGB camera getting sub-optimal overlap for photogrammetry purposes. We see a lot of 75/50 overlap which isn't going to work in any forested setting. The good news is the high quality IMU data gets into the image tags for using other methods of photo-based mapping.

TerrainDrop Posted at 9-20 12:18
The flight control for the L1 seems to optimize for the accuracy of the LiDAR data (as it should) but that leaves the RGB camera getting sub-optimal overlap for photogrammetry purposes. We see a lot of 75/50 overlap which isn't going to work in any forested setting. The good news is the high quality IMU data gets into the image tags for using other methods of photo-based mapping.

You can change overlap settings. But now i'm used to prepare two flight plan. One for Lidar, One for photogrametry. Then i let Terra managing the point cloud and Metashape managing the orthophoto. Sounds longer work but accurate.