Hello, my name is Guillermo Barros, I was using a trial version of pix4d, and trying to make a mosaic with images taken by a Sony camera. Longitudinal overlapping is about 80% and lateral overlapping is about 60%.
The images are JPG with EXIF data. The processing was made in a Windows 10 machine 64Bbit and 8GB RAM.
The orthomosaic, has some black stripes on it. Can this issue be due to low RAM or something like that.
I really apreciate your help.
Greetings.
Hi Guillermo,
Thank you for attaching the screenshots. Looking at it, I believe that the issue is not related to the lack of computational resources (RAM), but rather with wrong initial processing of the project.
Could you upload the quality report (…\project_name\1_initial\report\project_name_report.pdf) so I can get an overview of the project and processing options, here.
Best,
Hello, thank you for your answer, here i send link with the report files from the processing.
https://drive.google.com/open?id=1UPgcz7YrWEs14SWi3X8FGu9TIhwLGzSX
Hope you can give me some hint about what is going on.
Hi Guillermo,
There is a high difference in your camera optimization. Also, for homogenous projects like yours (eg: agriculture, fields), we recommend using the ag RGB template (it uses alternative calibration that is more suited for flat agricultural fields). The overalp image shows you had very less overlap (yellow and red portions). We suggest at least 85% frontal overlap and 70% side overlap which is missing in your flight.
Can you check if the camera parameters are okay? https://support.pix4d.com/hc/en-us/articles/202560899-Error-There-is-a-significant-difference-between-the-Initial-and-Optimized-Focal-Length. Also, from your next flights ensure you have enough overlap and flight height.
We will be glad to help in case you have more questions.
I have similar issues whenever I try to reduce gsd to minimize the file size. I don’t understand how the 1xgsd looks beautiful, then the 5 looks like this.
Matthew,
Can you try processing with global shutter? I have seen cases where using global shutter instead of rolling shutter removes the holes
I am Getting the same thing on my Ortho Mosaic, after the 1st step is ran the ortho preview that is generated looks perfect, when i do volumes for piles that looks perfect, my overlap is and has been at a 75/75 and i have never had a problem the last 20 times i have processed this exact flight, now all of a sudden my ortho has black spots in it. I have received this on my last 3 flights. In the quality report i show solid green for overlaps. Not sure what has changed
![Quality report1|690x410]Hi Chelsea,
This is, indeed, weird!
Could you please send us a screenshot of the orthomosaic and the processing options (from the quality report):
or better, you can attach the full quality report so that we can further investigate.
Thank you for reporting this,
Christina
PSY-01-29-20_report.pdf (3.3 MB)
Hi Chelsea, Sorry that this happened. Rolling shutter correction cannot cope with the arbitrary movements the drone has during the exposure of each image. We only model the linear rolling shutter and not arbitrary in our correction algorithm. This happens with the cameras like SQ RGB, Zenmuse, etc. They all have long read out times. Changing the setting to the global shutter as well as making the drone more stable or force it to stop when taking pictures should solve this issue (for the future flights). Do you think your piles are giving wrong volume measurements?
If i process the flight with the linear rolling shutter my pile volumes come back spot on, these are being cross checked by using a survey grade Leica GNSS survey system. I am also using GCP’s that have been surveyed in. Steps 1,2, & 3 process fine and everything comes in great other than the final Geotiff that i use to bring into Autocad. The preview geotiff looks great after step one has processed, the point cloud looks great for doing pile volumes, it is just the geotiff that i am having problems with. I have ran this several different ways and the only way that i can get the geotiff to come out looking normal is to change the camera setting from linear to global shutter, in turn this gives me inaccurate pile volumes. I have tried using 2 matches instead of 3, i have tried triangulation vs inverse distance weighted, i have tried selecting the advanced tab in the 1st step on the camera settings from all to all prior and also from standard to alternative as well among other things. It would also be very beneficial for me if in the desktop model, while using the volume selection, would come through in Cubic yards like the online version. Not sure why this has not been and update yet?
I am also not sure what has changed as to where the geotiffs are coming in different all of a sudden. My last 4 flights i have been able to achieve accurate data from processing but unable to come up with a decent Geotiff. These flights have been all different areas with different scenarios, ranging from heavy vegetation to sparse veg to open pits that we have mined.
As soon as the snow clears i will do some more test flights to double check the camera settings to ensure that those are correct, that is the only thing that i can think might be giving the processing a fit. I am using a Matrice 200 with a Zenmuse X5S with a 15mm focal length. It has a 35mm equivalent focal length and not sure if that got switched off in the Ground station Pro app that i use for flights. Curious if the Pix4d algorithms are trying to tweak something in there or what. I would assume that with the quality report showing that the whole flight area is green that i have enough overlap so i don’t think that those need bumped up. Please let me know your thoughts.
Since it is a rolling shutter camera, we recommend flying slow, Pix4D is not able to account for the arbritrary vibrations. I would recommend considering that for your future flights. For this flight you could do one thing, save the project using save as (after you do the volume and everything using rolling shutter correction). Save as will create a duplicate project. You can work with global shutter on the duplicate project to get the geotiff. The overlap is good, otherwise that would create an issue after step 1 in the quality report or the automatic tie points generated. Uisng more matches than less would create more accurate dense point clouds, it would remove all the noises.
"It would also be very beneficial for me if in the desktop model, while using the volume selection, would come through in Cubic yards like the online version. " I will surely suggest this to our team.
