I've done over 20 L1 flights with my M300 and mostly for contours, either in the city or in very wooded sites. We have a good workflow with benchruns using a surveyor level plus horizontal coordinates of the CGPs with a GNSS system, all using survey grade accuracy. The comparison results between GCPS and L1 data are within 0.10'-0.15' and I think this is pretty good for the L1 sensor.
But I've noticed recently that we have an issue with the curbs (or any similar structure). For example, this site we flew East to West and West to East, according to the path the drone takes when you plan the flight. I have concrete curbs on this site going in all directions, however the curbs perpendicular to the flight path (North-South or South-North) are deformed or stretched, it's shown on the picture below (Curb1), a horizontal distance between bottom and top of curb of 1.18' which in this case should be something around 0.3' it's a very big difference,  however the vertical distance (0.43') is very close to what is supposed to be.

On the other hand picture (Curb2), shows measurements more accurate to the dimensions of the curb, a horizontal distance from bottom to top of the curb of 0.33' and vertical distance 0.39'.

This pattern repeats through all my L1 lidar files. What comes to my mind is that DJITerra is stretching the curbs horizontally where the curbs are parallel to the flight path for some reason. I want to know if anyone has experienced that before ? If DJI has an update to fix that bug ? or if it's something with the way I'm flying?

My flight parameters are:
PPK
Altitude: 160'
Flight Speed 7m/s
Side overlap 50%
Echo mode Triple
Sampling Rate: 160KHz
Scanning mode Repetitive
RGB colouring ON
Gimbal angle: 90º (Nadir)

DJITerra Parameters:
Point Cloud Processing
Optimize Point Cloud ON
Smooth POint Cloud ON
Horizontal Datum WGS84
Geoid Default

Try with this :
PPK
Altitude : 260ft / 80m AGL
Side Overlap : 50%
Echo mode : Double (Triple is more for dense vegetation)
Sampling Rate : 80kHz
Scanning mode : Repetitive
RGB colouring : ON
Gimbal angle : 90

Then you need a point cloud post processing software, i use GlobalMapper, not the best but it does the job. Terra is only an interface from the drone to the LAS file. You will decimate your point cloud based on the tallest object, in this exemple 30m. The FOV is 70.4, with the settings I gave you above you will only keep what is included in an FOV of around 60. The goal is to remove all the parasites that are at the ends of the scan.

Dimensions are in meter.
Try also by unchecking the "optimize point cloud" function.
Otherwise there is software that does this very well and in an automated way, it's TerraSolid. Nothing to compare with DJI TERRA.

First I appreciate your opinion and I understand from the point of view of tall structures like the building you shown in the picture. But in this case I'm referring to a concrete curb, it's literately  6" above ground. With all respect I don't see why this would work elevating my altitude and reducing the number of shots per second (80KHz), which implicates in less points density.
I do work with a post-processing software besides DJITerra but that's not the case. what's I'm trying to show here is that the accuracy DJI promotes on Zenmuze L1 / DJITerra specifications its far from reality, unless I'm doing something wrong or my equipment is defective.

djiuser_RSLzfx2cRv7i Posted at 12-21 05:48
First I appreciate your opinion and I understand from the point of view of tall structures like the building you shown in the picture. But in this case I'm referring to a concrete curb, it's literately  6" above ground. With all respect I don't see why this would work elevating my altitude and reducing the number of shots per second (80KHz), which implicates in less points density.
I do work with a post-processing software besides DJITerra but that's not the case. what's I'm trying to show here is that the accuracy DJI promotes on Zenmuze L1 / DJITerra specifications its far from reality, unless I'm doing something wrong or my equipment is defective.

Yes sorry for my bad understanding of english.
Could you please share the point cloud extracted from DJI Terra with your settings.
I will compare on my data if i have such default. Few months ago my L1 started to provide strange data. The local dealer were not able to find the solution. I re-upload the firmware, reset the coloring from RGB at factory settings, and everything was and is still fine.

djiuser_RSLzfx2cRv7i Posted at 2022-12-21 05:48
First I appreciate your opinion and I understand from the point of view of tall structures like the building you shown in the picture. But in this case I'm referring to a concrete curb, it's literately  6" above ground. With all respect I don't see why this would work elevating my altitude and reducing the number of shots per second (80KHz), which implicates in less points density.
I do work with a post-processing software besides DJITerra but that's not the case. what's I'm trying to show here is that the accuracy DJI promotes on Zenmuze L1 / DJITerra specifications its far from reality, unless I'm doing something wrong or my equipment is defective.

Did you ever figure this out? I am getting similar results from tree tops. They are "elongated" in the direction of travel.

DJIUserL1 Posted at 7-4 12:26
Did you ever figure this out? I am getting similar results from tree tops. They are "elongated" in the direction of travel.

It is a matter of filtering. If you follow what i wrote above, you should be able to get rid of this trouble.
Do not use point outside of -30 / 30 degrees.

DJIUserL1 Posted at 7-4 12:26
Did you ever figure this out? I am getting similar results from tree tops. They are "elongated" in the direction of travel.

DJI doesn't have a solution for this, I went trough a 6 months+  back and forth emails, also sent my L1 to them be checked... No answers, no problem, they didn't even look at my unit. Anyways I've talk to other people with the same problem and the only reason I can think, its is because is a cheap lidar. The structure of the Laser in this unit (Livox) was not primarily made for this purpose, but more to work like a vision system for automated cars, it's literally a 500$ lidar unit. DJI sell for a higher price because of the technology embedded for the purpose the sell (IMU, RGB camera...) .
With that being said, now you know the limitations of it, you can work around the problem, like flying in a different direction to get better representation of the object.

djiuser_RSLzfx2cRv7i Posted at 7-13 10:22
DJI doesn't have a solution for this, I went trough a 6 months+  back and forth emails, also sent my L1 to them be checked... No answers, no problem, they didn't even look at my unit. Anyways I've talk to other people with the same problem and the only reason I can think, its is because is a cheap lidar. The structure of the Laser in this unit (Livox) was not primarily made for this purpose, but more to work like a vision system for automated cars, it's literally a 500$ lidar unit. DJI sell for a higher price because of the technology embedded for the purpose the sell (IMU, RGB camera...) .
With that being said, now you know the limitations of it, you can work around the problem, like flying in a different direction to get better representation of the object.

You talked to the wrong people. All LiDARs do the same thing. As the angle increases, the accuracy decreases.
It is necessary to remove the points located at the ends, they are unusable. This is why they recommend an overlap of at least 50%.

Before filtering I have the same problem as you. After that, the problem disappears completely.
Add to that the need to activate the IMU calibration systematically, I don't understand why Pilot2 gives us the choice. And preferably use a base station on site, no further than 10km, in PPK, not in RTK.

Feel free to send me a dataset to see if my method works and thus know if your unit is defective or not. I will be happy to give you a tutorial.

LV_Forestry Posted at 7-13 10:54
You talked to the wrong people. All LiDARs do the same thing. As the angle increases, the accuracy decreases.
It is necessary to remove the points located at the ends, they are unusable. This is why they recommend an overlap of at least 50%.

You clearly didn't understand the question. It's not about the side overlap, the features that get distorted happens when they have vertical change in the same direction(perpendicular) to the flight path. I've never had problem when the vertical change was parallel to flight path.

djiuser_RSLzfx2cRv7i Posted at 7-14 11:59
You clearly didn't understand the question. It's not about the side overlap, the features that get distorted happens when they have vertical change in the same direction(perpendicular) to the flight path. I've never had problem when the vertical change was parallel to flight path.

I offer you a simple thing. Share one of your problematic datasets, and we'll see.

Hey there! I stumbled upon your post about the DJI L1 CURB DEFORMED issue, and I can totally understand the frustration. Dealing with equipment malfunctions is never fun, especially when you're in the midst of a project. However, it sounds like you're handling it like a pro by considering pour concrete curb https://lilbubba.com/ instead.