Hi,

I recently bought a P4 multispectral. I am desperately trying to reconstruct images i took with my P4 but everytime i try it says "reconstruction error" at about 40% of the reconstruction process. I tried to reconstruct simple 2d maps (thus "non-multispectral") with rgb or NDVI pictures, and it works fine. but i need to perform a "2D multispectral reconstruction"...
Can someone help me?

Thanks in advance,
Regards

Hi,
Make sure you have activated your TERRA license if you have. If yes, can you give us more details, how do you proceed? TERRA version? Which files are you using ?  because TERRA is very simple, theoriticaly!

Make sure you have enough free space on the hard drive. And maybe try to reinstall the software if the problem persists, on a different computer if possible.
I proceed in this way: I create a new multispectral 2D project, then I choose to add a folder, I point to the folder where the photos are, and I click on start reconstruction. Simply.

I proceed in the exact same way (selecting a folder with multispectral images than clicking on "reconstruct"). I have already reconstructed normal 2d maps and 3d maps and everything works just fine.
The only difference is that i haven't yet paid for DJI Terra, i'm still with the trial version. Could that be the problem?
Otherwise I have a 32GB ram laptop with a Nvidia GTX 1050 Ti GPU (4gb ram). I have tried reinstalling but the problem remains the same. I have tried both versions of the software too (3.3.0 and the previous one). And i have a 1TB hard drive that is almost empty so that shouldnt be an issue.
I do notice that my CPU is running at 100% pretty much all the time... its an i5 processor. But thats okay according to what i read on internet...

Can you send me a file of your last flight with photos and GNSS RAW files? via Google Drive or otherwise. I will try to process them with my computer in this way we will rule out a failure of the drone or not, but I believe in it moderately. I have about the same PC configuration, the difference is that I have an i3 and it works very well.

Hi LV Forestry,
Thanks so much for helping me out.
Here is a link to a folder made by my P4 multispectral.
https://we.tl/t-mbfwbWYpMG
i'll send another one straight away, so you have two different datasets to try out.
Kind regards,

Here is the second folder, pictures were taken from 40m whereas in the first test they were taken from 10 meters i think with a very high overlap.
https://we.tl/t-6R4eHwFmEU

Arnaud LRC Posted at 2-20 23:53
Here is the second folder, pictures were taken from 40m whereas in the first test they were taken from 10 meters i think with a very high overlap.
https://we.tl/t-6R4eHwFmEU

Hi, can you send the images, the link has expired.

LV_Forestry Posted at 3-13 12:52
Hi, can you send the images, the link has expired.

Hi, I found the solution in the meantime.
I was trying to recontruct with multispectral images. The drone took one NDVI pic in JPEG and then 5 other pics in TIF (red, green, blue, nir and re). But you have to have a RGB picture (in JPEG) in order to obtain all the other indicators.

Arnaud LRC Posted at 3-13 23:41
Hi, I found the solution in the meantime.
I was trying to recontruct with multispectral images. The drone took one NDVI pic in JPEG and then 5 other pics in TIF (red, green, blue, nir and re). But you have to have a RGB picture (in JPEG) in order to obtain all the other indicators.

Thanks for this clarification - I received a similar error when I was trying to reconstruct some imagery that had an NDVI image followed by the TIF bands.

So how did you fix the problem?

Arnaud LRC Posted at 3-13 23:41
Hi, I found the solution in the meantime.
I was trying to recontruct with multispectral images. The drone took one NDVI pic in JPEG and then 5 other pics in TIF (red, green, blue, nir and re). But you have to have a RGB picture (in JPEG) in order to obtain all the other indicators.

What exactly did you do to fix the problem?

Jeramy Williams Posted at 3-24 13:14
What exactly did you do to fix the problem?

Well in ground station pro you have to be flying with the RGB camera selected. so when looking at what your camera sees in real time, it has to be RGB, not NDVI. If you are seeing with orange and green NDVI colors, you should go into settings, and change it back to "visual" if I remember correctly. Also make sure that all the other bands are activated. Because all these bands are needed to calculate the different indexes in DJI Terra.
I think DJI does this for people who only care about NDVI maps, not other indexes. and so upon arrival, you can quickly generate the Orthophoto with NDVI instead of uploading all the RGB, R, G, B, NI and RE pics into DJI Terra and waiting forever

Arnaud LRC Posted at 3-25 07:51
Well in ground station pro you have to be flying with the RGB camera selected. so when looking at what your camera sees in real time, it has to be RGB, not NDVI. If you are seeing with orange and green NDVI colors, you should go into settings, and change it back to "visual" if I remember correctly. Also make sure that all the other bands are activated. Because all these bands are needed to calculate the different indexes in DJI Terra.
I think DJI does this for people who only care about NDVI maps, not other indexes. and so upon arrival, you can quickly generate the Orthophoto with NDVI instead of uploading all the RGB, R, G, B, NI and RE pics into DJI Terra and waiting forever

When you use GSPpro, there is a button under your radio control, left side, which is used to switch from RGB to NDVI. The NDVI mode is very useful for quickly identifying variations in the index on crops, without having to reconstruct maps on Terra. After all depends on what you want to do, but if you want valuable results Terra is not enough.

i plan on making prescription maps to be used on a DJI T30 for variable rate application. not sure another program can do that. or is there?
also Terra for agriculture is 310 euros per year. most other programs are much more expensive

Hi

Below is a link which describes how to access the Terra log files and error codes for more detail on the underlying error, I hope this helps.
DJI Terra Error Code Guide

Arnaud LRC Posted at 2022-4-8 05:13
i plan on making prescription maps to be used on a DJI T30 for variable rate application. not sure another program can do that. or is there?
also Terra for agriculture is 310 euros per year. most other programs are much more expensive

Have a look at Open Drone Map
https://opendronemap.org/

It sounds like you found a solution, but also note that some stitching software will fail if you don't have all 5 bands for every images, so this can also cause a similar error message. this can be a hard one to debug because there is no way to tell if you do have all the images for each capture without writing code to pull the captureID out of the EXIF of each image (here's a code snippet that does this: https://gitlab.com/-/snippets/2493855)

Phenomics Posted at 2-5 21:41
It sounds like you found a solution, but also note that some stitching software will fail if you don't have all 5 bands for every images, so this can also cause a similar error message. this can be a hard one to debug because there is no way to tell if you do have all the images for each capture without writing code to pull the captureID out of the EXIF of each image (here's a code snippet that does this: https://gitlab.com/-/snippets/2493855)

No need to look at EXIFs. Simple arithmentic solution:

If your total number of  jpg + tif images is not evenly divisible by 6 you are missing images.

Or ignore the jpegs and tifs must be a multiple of 5.

This won't tell you if you have any corrupted image(s) but at least you'll know you have the correct number.

I give my trick to check that I have all the photos, even it never happened to have missing ones. I resize the explorer window to have 6 photos online. The JPG in RGB is always on the left. By scrolling to the bottom of the file, the JPG must be on the left. and There you go !



I use Agisoft Metashape. If a band or a few photos are missing, it does not make him hot or cold. It will reconstruct and propose the separated rasters at the end.

Phenomics Posted at 2-5 21:39
Have a look at Open Drone Map
https://opendronemap.org/

To date I have not found any software powerful enough to make prescription maps. Especially not WebODM.

Software that only asks you for an NDVI raster and outputs a prescription map is nothing more than eye candy.

You have to take into account the type of crop, seed, age, weather, elevation, irrigation...

If in the field there is a hollow where water accumulates, the NDVI will be very low, you will have thrown tons of fertilizer, apart from polluting the soil and squandering the budget of the farmer you will not do big thing.

Only a semi-automatic analysis with QGIS makes it possible to establish a prescription map, which includes the spreading of fertilizers/herbicides, drainage, irrigation.

There is a significant amount of scientific literature that demonstrates that the NDVI/fertilizer relationship is not as simple as the algorithms of these miracle software would have us believe.

LV_Forestry Posted at 2-6 10:58
I give my trick to check that I have all the photos, even it never happened to have missing ones. I resize the explorer window to have 6 photos online. The JPG in RGB is always on the left. By scrolling to the bottom of the file, the JPG must be on the left. and There you go !

[view_image]

That works too

LV_Forestry Posted at 2-6 10:58
I give my trick to check that I have all the photos, even it never happened to have missing ones. I resize the explorer window to have 6 photos online. The JPG in RGB is always on the left. By scrolling to the bottom of the file, the JPG must be on the left. and There you go !

[view_image]

That's a good trick!

patiam Posted at 2-6 10:37
No need to look at EXIFs. Simple arithmentic solution:

If your total number of  jpg + tif images is not evenly divisible by 6 you are missing images.

That's fine but then if you are missing a few images in the middle of the flights there's no easy way to tell which is the capture that is missing (other than the handy folder trick that LV_forestry posted).

Phenomics Posted at 2-6 15:58
That's fine but then if you are missing a few images in the middle of the flights there's no easy way to tell which is the capture that is missing (other than the handy folder trick that LV_forestry posted).

You said there's no way to tell if you have missing images. LV_Forestry & I gave you two ways.
Now the goalposts have moved to finding out which images are missing. That's simple too:
The images follow a simple numbering schema wherein the last character of the filename is the digit 1-5, denoting band #. Simple searches with no scripting can reveal which band(s) may have missing images. Then one can figure out which captures are the offending ones. If one has some skill with python or powershell, this can be done much more efficiently. But no need to go into the exif information. My point is just that there's more than one way to answer the question.

And, as LV_Forestry pointed out, if you're using a halfway decent photogrammetry application, it iwll be tolerant of captures with missing bands.

Phenomics Posted at 2-6 15:58
That's fine but then if you are missing a few images in the middle of the flights there's no easy way to tell which is the capture that is missing (other than the handy folder trick that LV_forestry posted).

Having missing images is very strange. Check that you do not exceed the minimum interval for saving images. Clearly flight slower or higher or with less frontal overlap.

Then I join the words of Patiam, I imagine that you use DJI Terra. It is a very capricious and very intolerant software.

LV_Forestry Posted at 2-6 10:58
I give my trick to check that I have all the photos, even it never happened to have missing ones. I resize the explorer window to have 6 photos online. The JPG in RGB is always on the left. By scrolling to the bottom of the file, the JPG must be on the left. and There you go !

[view_image]

I should add that I have a colleague that uses the same method to select and move all of the thermal IR band images for his Altum data (to save time/space/quota/etc) b/c he does not care about them. Size the window correctly, select the last band, and execute. It works.

Hi,
I had the same problem, but with a mavic 3M drone. The output of this device is not 6, but 5 images (no blue channel between the TIF images).

2D and 3D images will do the terra, but only in visible light mode, it throws an error in multispectral.

Has anyone encountered this problem, thanks in advance for any replies.