Goodmorning everyone.
Today i'm asking you for some advices on surveying a vertical rock wall with the DJI L1.

It's not the first time i need to survey a vertical wall, but it's the first time i need to survey a HUGE vertical wall, which is about 200ha surface. For small vertical surfaces i used manual flight without any problem, but in this case, where we have a 3km lenght and in the highest point almost 900m in height differences, this operation needs to be as precise (and planned) as possible.
Here's a screenshot of the area


I divided in 5 macroareas, each one with a different take off point, to respect regulations. The lidar flight should be about 100m in distance from the terrain and here's the first problem.
I have the DTM of the area, and if i create a MAPPING MISSION, i could use it to provide a Terrain Follow mission.

But terrain follow mission puts the camera in perpendicular to the terrain, and would not provide best data accuracies for a vertical mapping survey. What i need (and what i've ever done in manual mission) is a mission where i have the lidar perpendicular to the vertical rocks.

so my question is: is it possible to createa an automatic flight plan and take manual control of the gimbal?

I tried moving the gimbal while the matrice is processing the mission, but every waypoint it re-put the camera back to default position (zenital). I tried to create a manual waypoint mission (which is possible to set the gimbal to manual control), but for a such big area with complex terrain heights is more complicated to plan (on the smart controller.... at least i could do it on pc!!) rather than to flight manually.

The way i would operate will be: pre-plan a mission BUT not starting it, just to have it the flight paths on the map, and try to follow it manually. Is it possible to view the Argumented Reality waypoints in the screen without starting a task?

Any other advice for an automated flight for this kind of area are welcome!

Hi Guys!
First day covered, i've done all the red area, the left side of the blue, and the lower part of the orange area, in total about 75 hectares with 3 batteries (3x2 obviosuly), where the first half i forgot to calibrate before scanning, so i restarted scanning but about 45% battery gone...
All manual flight, path perpendicular to countours line.
The auto reconstruction with the optimization in DJI Terra PRO made an incredible work. Once finished the rest of the survey, i'm going to correct the trayectories in TerraMatch with some ground control points, i'll show the results.
But till now, this reconstruction with all manual flights exceedes my expectation. I thought I would have to work on it a lot in terramatch to align the scans well.
This is some screenshot of the preview. I've imported in Spatix to look at the alignment, it's simply perfect (i don't have screenshot yet).

Bravo Panph,

How did you make your manual trajectories? from bottom to top or from left to right. Do you choose a timelapse function for the pictures?
Yesterday I was in the Alps for a railroad tracking mission. It's stressful to fly the Matrice in a narrow valley.
Have a good flight.
JL

Aaaand, we did it!!
9 flights, about 4 hours of scanning, 1,35 BILLION points, and all the mountain is ready.



This have been an hard job, specially for the high winds on the lake, maximum at 12m/s, but the Matrice held it pretty well!

Fans3cf the trayectories was from bottom to top, perpendicular to countour lines, Lidar from 30° to 70° depending on the slopes, on vegetation 90° to ensure maximum coverage with third return. The pictures were take with fixed time interval, because i also want to work with them in Pix4d (photogrammetry) and see if i could get a discrete orthomosaic from them, but was not required from the customer.
The model is complete, i have to delete lots of point to make it "workable", because at the moment, 1.3 billion points could not run on my pc (i9 14core, 2080ti, 64gb ram).



Here you can see different color for different FILES (files, not paths), and two sections. From top to bottom is about 850m on highest point.

Bravo Panph and thank you for this feedback.
The Matrice300 is a great machine. I wish you all the best.
JL

Hi Panph,

congratulations !!!  You did a great job considering the complex  topography of the site and that you manually piloted the M300RTK   :-)

At present we are using DJI Zenmuse L1 for surveying some steep slopes and cliffs on the Island of Ischia (southern Italy) for geomorphological  studies.

For this purpose we are planning our missions mainly  using SPH UgCS software which allows to do Terrain Following and to setup many flight and some camera parameters (e.g. gimbal tilt/pitch). It does not allow automatic autocalibration every 100 sec. but you can eventually insert in your flight plan an intermediate calibration pattern (i.e. eight or U maneuver)

DJI Pilot is more powerfull and gives you more control both on the camera and on IMU automatic calibration when flying at fixed height AGL, but  such autocalibration feature is deactivated if you choose a  Terrain Following  LiDAR mission.

About cliffs I suggest you, when possible, to do both nadiral (90 degrees) and  high oblique (perpendicular to the slope) surveys to maximise LiDAR point cloud coverage behind obstacles along the cliff.

We usually plan flight paths parallel to the contour lines to be able to use flight speeds higher than the M300RTK maximum vertical speed, but this is not a must.

We are also experimenting the Vertical Scan route type in UgCS.


I hope that others will share here their experience on using DJI Zenmuse L1 on steep slopes and clifs.

Buon lavoro,

Maurizio


P.S.
For italian speaking pilots I suggest to visit the 3D Metrica website (by Paolo Corradeghini)

mbuonanno Posted at 3-30 08:55
Hi Panph,

congratulations !!!  You did a great job considering the complex  topography of the site and that you manually piloted the M300RTK   :-)

Grazie Maurizio!!

thank you for your advices. I've done exactly as you said, doing both nadiral and oblique flights, but i've done nadiral flights only where vegetation was pretty dense to let the laser hit easly the ground.
I am amazed at how dji terra has aligned the manual flights, considering that some parts I have not flown following a precise path.

To have a path as linear as possible, I created the mapping missions, considering terrain follow and overlap at 35%, then I left them open below in the map and I followed the paths manually. In some areas, however, I flew manually for better results, turning and changing altitude "as i want", but DJI Terra PRO managed it pretty well.

But doing flights parallel to contour don't you have errors in alignment due to different "error" from close or distance points? i mean, at the highest point you'll have points with low errors, but in the lower part you'll have a more sparse cloud to align. As in photogrammetry, i plan flights perpendicular to coutours to let the GSD to be much costant as i can, to avoid "resolution errors".

I wanted to try UGCS, but unfortunately my mobile devices are all Apple, so i cannot install UGCS on my laptops/tablet. I should have bought a dedicated one.
I hope DJI, maybe with the new Pilot2 app would solve this kinda problems, being the matrice300 solution the flagship of their fleet (or do we have to wait [and buy] the new 400?).

Anyway the work result are way better than i expected, pretty excited!

Grazie Maurizio per l'imbeccata su 3dmetrica, lo conoscevo in quanto lo consultano miei colleghi, ma magari ora ci faccio un bel salto. Ho visto alcuni video su youtube di Corradeghini, sicuramente un riferimento!

Buoni voli!
Alessio

Panph Posted at 3-30 10:41
Grazie Maurizio!!

thank you for your advices. I've done exactly as you said, doing both nadiral and oblique flights, but i've done nadiral flights only where vegetation was pretty dense to let the laser hit easly the ground.

Hi Alessio,

you said:  "But doing flights parallel to contour don't you have errors in alignment due to different "error" from close or distance points? I mean, at the highest point you'll have points with low errors, but in the lower part you'll have a more sparse cloud to align. As in photogrammetry, I plan flights perpendicular to contours to let the GSD to be much constant as I can, to avoid "resolution errors"

I agree with you that  when using photogrammetric Structure from Motion software it is better to keep GSD as constant as possible to avoid problems matching homologous points.

When the slope is VERY STEEP (e.g. a cliff) clearly the best is to fly  with the camera oriented perpendicular to the face, doesn't matter if paths are vertically or horizontally oriented. Nevertheless, it is advisable to add a nadiral coverage to get pixels behind vertical obstacles along the cliff.

When surveying STEEP slopes with nadiral photographs I agree with you that "terrain following" and paths about perpendicular to the contours reduce issues related with changes in scale/GSD. Higher flight height AGL can eventually reduce issues due to paths "parallel" to the contours.

When doing LiDAR surveys we do not have the GSD constraint but on the other hand we have to take into account that LiDAR point accuracy decreases with both scanning angle and distance from the emitting source. This gives you much more flexibility, and allows you to plan a high accuracy mission or a lower accuracy mission if you need to describe the landforms in less time.

Again, When the slope is VERY STEEP (e.g. a cliff) the best is to fly  with the sensor oriented perpendicular to the face, doesn't matter if paths are vertically or horizontally oriented. As you said it is advisable to add a nadiral coverage to get LiDAR point behind vertical obstacles and more points on the ground.

When doing nadiral surveys on STEEP slopes with a low cost LiDAR sensor (like DJI Zenmuse L1), flying along paths "parallel" with the contours you will certainly get asymmetric accuracy errors across the path line. But if your main target is not accuracy I believe you can increase lateral overlap and postprocess LiDAR data to use only the points closer to the flight center line.

Anyway I am not a LiDAR expert and we did no organize a rigorous test yet to quantify what I wrote. Consequently it would be very useful to share further experiences with you and other users. We could start sharing the above topic with Paolo Corradeghini.  ;-)

Buon lavoro,

Maurizio


P.S.
A nice article in italian: Un rilievo fatto con un LiDAR montato su un drone | 3DMetrica

Panph Posted at 3-30 10:41
Grazie Maurizio!!

thank you for your advices. I've done exactly as you said, doing both nadiral and oblique flights, but i've done nadiral flights only where vegetation was pretty dense to let the laser hit easly the ground.

Hey Alessio, so did you use the DJI Pilot 2 app for your mission planning on the rock face? Would you mind sharing what mission type and settings you had in place for that scan? Specifically interested to know how you got the 90 degree scan. We are conducting a similar flight and would love to use Pilot 2 instead of UgCS if possible.

WS-Evan Posted at 9-28 13:05
Hey Alessio, so did you use the DJI Pilot 2 app for your mission planning on the rock face? Would you mind sharing what mission type and settings you had in place for that scan? Specifically interested to know how you got the 90 degree scan. We are conducting a similar flight and would love to use Pilot 2 instead of UgCS if possible.

Hi Evan,
yes i've managed to do it with pilot2, but it was very complex and difficult to program. It seems impossible that a company like DJI can't (or won't) develop a versatile app for its products.
Anyway I performed the missions in 2 different ways, starting from the same base.
to begin with I created a mission as a mapping, with correct overlap, as if it were a flat area, in order to see the horizontal positions of the drone path.
I have saved this mission, exported and displayed on a pc monitor.

From here I created 2 types of missions, based on the complexity of the terrain. The most "simple" missions, that is where the rock was not extremely vertical, I created waypoint missions, with horizontal position as the mapping, but where each waypoint I assigned an altitude verified on google earth 3d (terrain + 90m approximately), I left the gimbal in manual mode and then during the flight I set the gimbal to "free position" instead of "follow aircraft heading", in order to position it as perpendicular as possible to the rock to be scanned, and the drone was free for moving.

I carried out the more complex portions in manual mode, but leaving the "mapping" mission open so as to see the traces in the map at the bottom left of the screen. Always free gimbal mode of the L1 and trying, looking at the point cloud, to obtain the desired overlap (about 35%). It would have been useful to have a measurement of the distance to the center point on the screen, so as to always understand the drone / ground distance and keep it as constant as possible, instead of hoping to do the right job.

ottimo filo guys!