No match between nadir and oblique aerial photography

[dmct]

Hello guys!

I´ve been using photoscan for a short while and must confess although I´m getting the hang of it, I am no expert so my question might be a little essential.

I´m working on monitoring changes in semistable dunes. For my purposses I generate DEMs and orothophotography with Photoscan Pro using aerial photography taken with drones. Recently I encountered a problem when aligning a chunk. I always group nadir and oblique frames in the same chunk in order to align them toghether and get the most accuracy posible, but in this chunk I am getting no alignment with the oblique imagery. When analizing matches I noticed oblique frames have no valid tie points where there should be. My parameters are the highest accuracy enabling generic preseleccion and am using 40 000 and key point limit and 4 000 tie point limit. I am also masking some of my oblique imagery since I am getting some background wich I am not interested in.

I tried disabling reference information and realigning using only GCP but I am getting the same results.

I include two screenshots, one of nadir and one of an oblique frame of the same area where a GCP is visible.

I have generated numerous projects of this same area using the same flight plan in different dates, but I am only having trouble with this project. I would really appreciate if you could help figure out why.

[SAV]

Hi dmct,

Here a few things you should/could try:

1. For the image alignment step use Pair preselection: REFERENCE. UAV imagery is generally geotagged (location is stored in each image). By using REFERENCE this info is used to assist the photo alignment step.

2. Also use your GCPs during the initial photo alignment step. Make sure that you mark them in both nadir and oblique imagery. This will help PhotoScan to match pictures. You can also add more markers without any known location which PhotoScan can use as 'tie points' to improve image alignment.

I hope that works.

Regards,
SAV

[dmct]

Hello SAV,

first of all thank you for the tips, it worked halfways, I managed to get valid tie points in oblique imagery but it is still not matching oblique with nadir frames. Not only that, I am getting different depths in my sparse cloud, one corresponding to oblique frames and another to nadir.

Any idea of why this is happening?

[SAV]

Hi dmct,

I assume you are using a drone from DJI. There is a known issue/problem when it comes to altitude/elevation information from geotagged images that have been taken with drone from DJI (check this forum). It does not record the elevation above sea level (as you use for your ground control points), but the altitude (height) above the point of take off.

If, for example, you have several surveys with different points of take-off, then PhotoScan might struggle to match these imagery based on the reference information.

Therefore, you have to either fix the altitude information in the geotagged images or try to adjust the Z accuracy (default is 10m) for your images in the reference settings.

Note that matching results of nadir imagery with oblique imagery improve if you have sets of oblique imagery at different levels with different angels. For example, one flight line at 10m above the ground, camera at 65 deg from horizontal and another flight line at 20m above the ground with the camera at 75 deg from horizontal.

All the best.

Regards,
SAV

Hello SAV,

first of all thank you for the tips, it worked halfways, I managed to get valid tie points in oblique imagery but it is still not matching oblique with nadir frames. Not only that, I am getting different depths in my sparse cloud, one corresponding to oblique frames and another to nadir.

Any idea of why this is happening?

[dmct]

Hello again SAV,

Indeed, I am currently flying with Inspire1 from DJI. I was under the idea that the altitude recorded was above sea level, but I´ll have to check that since I am using a post processing RTK to geotag my GCPs. If what you say is true, that  would possibly be the main issue. It is weird though since I used the same GCP´s in all other projects and had no problem.

Thanks a lot for the useful information!

Regards
dmct

[cadm8]

I agree with SAV, you ideally need a well pre-planned flight scheme going from nadir to oblique imagery and various altitudes. I have noticed that irregularly or randomly captured photos can produce photo alignments with false elevations

[SAV]

Hi dmct,

I've had the same problem about 2 years ago using a DJI Inspire 1 too. My workaround was to also measure the elevation above mean sea level at the point of take off (using the same RTK GPS I used to measure the GCPs) and then add it to the altitude above ground level value that is recorded by the drone and stored in the geotagged images.

Alternatively, as I already mentioned, you can choose a very low accuracy for the Z coordinate of geotagged images which will almost neglect the impact of the altitude value.

Regards,
SAV

Hello again SAV,

Indeed, I am currently flying with Inspire1 from DJI. I was under the idea that the altitude recorded was above sea level, but I´ll have to check that since I am using a post processing RTK to geotag my GCPs. If what you say is true, that  would possibly be the main issue. It is weird though since I used the same GCP´s in all other projects and had no problem.

Thanks a lot for the useful information!

Regards
dmct