Hello,

I use the M300 and the L1. In the point cloud I always notice these tracks. They remind of the flight path. They are also visible in further processing. I have tested various parameters, calibrated the L1 and researched a lot, but unfortunately I have no solution for this. Do such tracks also appear in your point clouds? Is there a solution or is there something wrong with my camera?

Greetings
Sandy06

I'm a complete noob with Photogrammetry and 3D modelling in general. I've recently started trying to learn photogrammetry and am particularly interested with importing point clouds created in Agisoft Metashape (.las files) into UnrealEngine (or other software) with the goal of making animated videos. I've been teaching myself with some help from friends and random tutorials/online documentation.

Kodi nox

Hi Sandy06, we have the same issue and no solution yet. It's a typical issue of the L1. I've contacted DJI support and still waiting for an answer.  If ever you found a solution please share. Thank you.

DOM_7806 Posted at 4-18 03:54
Hi Sandy06, we have the same issue and no solution yet. It's a typical issue of the L1. I've contacted DJI support and still waiting for an answer.  If ever you found a solution please share. Thank you.

Hello DOM_7806,

how do you know this is a typical L1 problem? I appreciate your message, thank you! I was not sure if this problem also occurs with others. I have written with DJI support and implemented all of their suggested solutions, but to no success. Now they have asked me to send the camera in for repair. I have not received any confirmation from DJI that they are aware of this error. I have noticed, that DJI is often very well aware of problems, but does not disclose them to subsequently clean up with updates or similar. This assumption I have read a few times in the forum, as well as recognized myself. As soon as I get my device back I will report again.

Greetings,

Sandy06

Hi Sandy.
Are this "track" only zones with superior density of points or these points are at different heights than the "surface around"?
Which scan path did you choose when having this issue?

If it's only a different points density, it might not be a problem. If you're using non-rep scan path it has a more dense area in the middle. When the L1 is perfectly nadiral to a surface, and this surface have an high index of reflectivity, the points on the ground could only reflect just in the middle of the scan path. The extreme example is when scanning a lake or simply water, L1 register points only at the center of the scanning path.

But i've never experienced "lines" like yours. I'd like to know what you were scanning, and if other points of the surface are in "right density" or if density are too low, because i see a lot of holes in your scan.

Hello Panph,

these traces occur across the entire point cloud. They are always along the flight path.

I collect the data to create DSM. There is always lots of vegetation, water and sometimes buildings. Vegetation makes up the largest portion of this. Water, i.e. rivers, only a very small part. Especially with straight areas, like a field, I noticed more and more errors in the elevation model (Screenshot). Finally, I looked at these places in the point cloud and realized that exactly along the path where this point density changes, the errors occur. I have attached a screenshot of such point density difference. I used a Path Profile tool to create it. Overall, these points are then at least 30cm higher or lower than the rest of the surface.

I never use the non-rep scan path. According to my research and also DJI's recommendation, I always use rep scan path.
These traces do not occur over the entire point cloud. They are randomly scattered across the point cloud. I don't see a pattern to this, so I could say that these errors always appear only at the edge, for example.
I do not think that I have a too low point density. Please note that I zoomed in very close in the first screenshot. However, I have a point density of 1600 samples/m2. I think that this must be sufficient.

ok. thank you for your explaination.
I could assume this errors, as you said you have it at the edge of the area, are due to Rep-scan and M300 turning. While turning the laser can create this kinda zones with more density but less accuracy, because the L1 have lots of movement to deal with.
I've experienced this "errors" only with rep-scan and manual flight, because i use to delete "turning points" in terrascan if i'm surveying with automatic flights.

To delete this parts, in Terrascan you can delete the parts of trayectories you don't need. In the wizard, it could automatically delete the turning trayectories (with the "low accuracy" points), and even the short trayectories (the parts between 2 "main trayectories").

IMHO your model could be more reliable with this procedure. Do you use the point cloud optimization in DJI Terra?

Hello Panph!
Thank you for your reply. I am sorry that I am just now replying. The tracks seem to be related to the movement of the M300, I agree. However, I always fly automatically as well and the tracks then move through the whole area so there is a clear pattern. However, I have not yet found a solution to adjust or delete the points whose height is not correct. Do you have a solution for this or is that possible with Terrascan?

Yes, I always use the cloud optization in DJI Terra.

Greetings Sandy06

Hello Sandy06,
I have been using the L1 since October and have noticed and dealt with the problem you describe. The problem is that the L1 is very inaccurate along the edges of the swath/flight line. If you are using a software like Terrascan you can split your trajectory at the turns and end up with each flight line separate from all the others with their own trajectory. Once that is done you will notice that these areas of mismatch are all along the outer most edges of the flight lines. I have dealt with this by filtering out the lidar points from the edges and only using 60 degrees of the field of view. There is a routine within Terrascan to perform this ALMOST automatically called "classify by angle" in which you can classify anything outside of the set range into a point class that you can exclude from your ground points or models. This has greatly increased the accuracy of the data.  See the screen shots below.

Hello Jesse G,

thank you for your answer. I have read several posts in the forum that describe this error roughly. Therefore I know that the problem exists. But you are the first one with a concrete solution. You describe exactly my problem. I will try to get a trial version of Terrascan and test if this works. Your photos look very promising. Do you also have a solution to get the data already in good condition? So that you don't have to post-process them with another software again? Some different flight settings or something?

I have never used a Lidar sensor before the L1 and I don't know anyone who has one. However, I find it extremely strange that I have to buy another software to get my data the way I need it. Do you know if this process is standard with Lidar data, or is that a clear mistake on the part of the L1? In my opinion this should not be the case. I would understand if sometimes a dataset has this error, but these errors are found in all of my dataset

Greetings Sandy06

I also recommend Terrascan's cleaning algorithms.
A question:
How much side overlap have you set up? If I remember correctly, by default it is set to 20%, but personally I do not go below 50%

aebrilievi Posted at 5-24 06:14
I also recommend Terrascan's cleaning algorithms.
A question:
How much side overlap have you set up? If I remember correctly, by default it is set to 20%, but personally I do not go below 50%

Hello aebrilievi,

thanks for your input.
I always use 50% side overlap.

Greetings Sandy06