Dense cloud generation not working, Zero Resolution error message.

[MJoelle]

Hey!

I am processing thermal imagery in Agisoft 1.8.3. Pictures were taken with XT 640 x 512 de FLIR. I found pixel sizes were 0.017*0.017 (in mm).

My first data set was taken in 2018 over a lake. Processsing worked great and was able to put together a good thermal orthomosaic.
1. I import my picture folder and select Multicamera system. 2. I calibrated the camera with pixel size. 3. Aligned pictures. 4. built  densecloud 5. DEM and 6. Ortho for every chunks. 7. Applied raster transform with formula (B4/100)-273.15 to get degrees in celcius.
I aligned chunks with the RGB orthomosaic (marker based) and merged thermal chunks aftewards. It all worked great. 

My second set of data is with the same drone and camera, over the same lake, but in 2019. I input same camera parametres and aligned pictures, which worked well. My Tie point cloud is looking good. But when I ask the software to build the Dense Cloud, it doesnt work. It only partially buids it (small scattered areas) and depth maps are minimal (4 to 10 for 300+ photos). It also sometimes shows error message saying Zero Resolution and generates nothing.

Is it possible that my pictures were corrupted if they were in a zipped file (RJPEG pictures)? First set of data was not shared in a zip file.
What am I missing. Why is the processing for same type of data not working now?

Any leads to what I should look into? Thanks a lot!!

[Paulo]

Hello Mjoelle,

the zero resolution may be caused by misplaced reconstruction volume see.

Zero resolution -  Bounding box is misplaced prior to the Build Dense Cloud operation and doesn't contain any information (almost no tie points are present in the bounding box and therefore no depth maps can be generated).
Sometimes this message appears after the depth maps calculation, if all the generated depth maps are empty - such behavior has been observed on old AMD graphic cards with out-dated drivers. In similar cases it is recommended to install the latest compatible GPU drivers from the official web-site and rebuild the depth maps (disable "reuse depth maps" option in Build Dense Cloud / Build Mesh dialog).

Also could do a print screen of your tie point Show info window to get better idea...

[MJoelle]

Hi Paulo, thanks for your quick answer!

My bounding box seems fine from what I know. Most of my tie points are in. It doesn not differ from last processing (2018 dataset).

Here are some screenshots of Infos. You can also see what happens with dense cloud generation.

Appreciate your help!

MJ

[Alexey Pasumansky]

Hello MJ,

Do you use any GPU for the processing? If any is enabled, please specify its model and related driver version.
The problem could be related to the outdated AMD driver, if you are using AMD GPU.

[Paulo]

Hello MJ,

what amazes me with your alignment is the RMS reprojection error of 0.088 pix. It is incredibly precise. Also you have something like 4000 projections per image which is also considerable  given 640x512 pixel imagery (albeit oversampled).... Quite surprising!

[MJoelle]

Hi Alexey,

I am not very familiar with GPU, but from what I see, I am using NVIDIA Quadro P620, driver version being : 30.0.15.1215.


What is strange to me is I am able to process the same type of data and create good dense clouds and depth maps (hence good DEM and Orthomosaic) from Data Sets that are very similar (same overlap of 90%, same resolution, same thermal camera, etc..). I linked a screenshot of successful processing with first dataset, and unsuccessful processing with second data set, over the same lake area with same amount of pictures.  Both processes are using same GPU...

Thanks for your help again, and hoping to find a solution for this.


MJ

[MJoelle]

Hi again Alexey and Paulo,

Perhaps Aggressive DepthMap filtering mode for my second data set is not adequate since it seems to be covering more of the lake area and less of the topography (hills on either wide of the lake).

I tried with Mild filtering and it seems to be a bit better (no more Zero resolution report, but still scattered DCloud and minimal Depth Maps). Still, I am able to put together an orthomosaic. Now, will it alter my thermal data if depth maps and Dcloud are minimal? Or is it strickly aftecting my elevation data (which I don't really need)?

Thanks again 

[aarondesalvio]

Hi again Alexey and Paulo,

Perhaps Aggressive DepthMap filtering mode for my second data set is not adequate since it seems to be covering more of the lake area and less of the topography (hills on either wide of the lake).

I tried with Mild filtering and it seems to be a bit better (no more Zero resolution report, but still scattered DCloud and minimal Depth Maps). Still, I am able to put together an orthomosaic. Now, will it alter my thermal data if depth maps and Dcloud are minimal? Or is it strickly aftecting my elevation data (which I don't really need)?

Thanks again 

Hi MJoelle, according to this reply by Alexey, dense cloud quality does not affect the resolution of the orthomosaic. Hope this helps!

https://www.agisoft.com/forum/index.php?topic=12247.0

[schahied]

Hi All. I am having the same issues as discussed below. When i go the driver options my computer says they are the latest one.

[Alexey Pasumansky]

Hello schahied,

Can you please share the processing log related to the operation?
Please also check that the bounding box (region) in the Model view includes the area of interest.