Hi all, hoping someone can give some tips with this. I have been struggling to calibrate a dataset of a coastline where around half of the images are over shallow sea. I need to keep the sea section in as we’re mapping a seagrass bed, so the loss of it makes the dataset worthless. I had struggled with this in other programs but an acquaintance had run the dataset in Pix4d fields and produced an orthomosaic which had most of the sea section in which convinced me to go ahead and buy Pix4dmatic as the description of Pix4dMatic sounded like it was the best option. I’ve read the documentation and tried to change scales and template (although primarily used ‘Flat scene low texture’) but none of it seems to be improving the area that can be calibrated.
I don’t understand why it’d work so easily in Pix4d fields yet seemingly be impossible in Pix4d matic so i’m hoping someone has some advice on this.
cheers, Peter
Pix4D Fields creates a lower quality ortho map. I would bet that Pix4Dreact would also work and also create a low quality orthomap.
Is the water moving (waves)?
If so I would not trust the outputs if they calibrated anyway.
I map beaches for costal erosion and all of the water sections are untrustworhty garbage. Moving waves, reflections and other water issues will affect any calibration. If you research this you can find some coastal studies over clear water had similar problems.
Sorry for the delay in replying to this. The water has small waves but no white water. We’d aimed for calm seas but not really an option most of the time!
Would there be a way to create a lower quality ortho map within pix4dmatic or the packages are just operating too differently?
The approaches are different, the one in react/fields is more about “stitching” whereas the one in matic (at this point in time) is based on a different approach. The stitching one is more adapted to this kind of projects. We may integrate that algorithm in PIX4Dmatic in the future to solve the kind of issues you are facing directly in PIX4Dmatic too.
That said, I’m not sure that “flat scene low texture” is the right way to go for a coastline in PIX4Dmatic, that’s usually more for flat surfaces in agriculture. Have the other templates not worked either?
In terms of kelp, I can confirm that someone else managed to do this with PIX4Dreact, you can read more here: Overwater mapping: the case of the disappearing kelp
@Pierangelo_Rothenbuhler thanks for this explanation. Frustrating to find this now after purchasing the perpetual license with the aim of processing this type of dataset, especially when Matic is . Currently none of the templates have worked. I keep experimenting, reducing scales etc but so far it’s all been pretty disappointing. I was using flat scene, low texture as a sandy/gravely beach seemed to fit that description well.
Do you have a timeline for when the algorithm may be included in Pix4dMatic?
Many thanks,
That’s unfortunate, sorry about that. That said it’s a rather challenging use case, I do not have a clear date on when this algorithm will be part of PIX4Dmatic, for the moment it’s not planned to have it this year. What I’d suggest is that you try out PIX4Dreact and check whether it works for this case. Can you let me know how it goes? Then we’ll go from there.
Hi Pierangelo, Thanks for this advice, i’ve downloaded the Pix4dreact trial and it processed this well. It’s got some small gaps but great coverage for the full extent of the seagrass beds. So it’d be good to know how best to proceed with the fix- Would this involve the purchase of Pix4dreact?
Additionally to this problem is another large piece of our photogrammetry work we have has been trying to create photogrammetry mosaics from underwater video (similar to this conversation - Importing Photos Extracted From Video). Again I hadn’t really understood the differentiation of the different Pix4d packages and thought Pix4dMatic would cover these needs. I have been attempting to create the correct Metadata using EXIF tool and ffmpeg for extracted frames but not got there yet. It may be worth having a quick call at this point if that’s possible?