Richo Theta V - ray cloud artifacts

Hi guys, I have taken several photos outdoors today with decent lighting (daylight). When I process the photos I am getting a lot of stray points in places where the surfaces should be flat. (building planes and road). I am also not getting a very dense ray cloud despite the settings. 

I have some questions, could this be attributed to the 12MP and sensor of the camera (not enough pixel information). 

I wanted to incorporate a camera to help myself capture more of the building facades and other smaller areas of interested that a sUAS can not get into. 

Here is the project screen shot.
I had something similar happen a few years ago when using a parrot bebop (v1).

Am I better off just using a non-spherical camera and just take more photos?

 

Looks like you need more pictures to me but there are other variables as well…easiest to start with more pictures first.

Hi Adam, I took several (more than 20 or so).  I am wondering if using spherical photos is the culprit.

I am about to go take several photos with my Samsung Note 8 in a few minutes to compare.

It would be great if the pix4d team could confirm.

PS: Is there an easy way to remove the sky from several photos all at once, I seem to be getting quite a bit of stay points for the sky.

Here is what the current elevation looks like:

Good Morning everyone, quick update.
I took these photos with a Samsung Note 8. The point was processed at low quality and 2 matches per pic.

You can clearly see the improvement on the details that were picked up by the software. Yes maybe I took more photos this time but more to account for the fact they are flat pictures as opposed to 360. Photos were taken in both projects at approximately the same distance, both from the building and each other.

This project in the bottom seems to have developed much more detail Camera specs are 12MP it is still a cell phone sensor  1/2.5"

I am really thinking that the poor cloud resolution on the Theta V data set is attributed to the lower megapixel count along a 360 photo. Which makes the software struggle. I experienced this when I tested the software with a bebop a few years ago. That particular UAS takes 12 mp spherical photos only. The results were very similar.  

Pix4D any thoughts?

The Theta Camera seems to work best for building interior photos.  When used outside, there can be a large amount of pixels within  each photo that ultimately captures sky and distant background objects (doesn’t contribute to the reconstruction of the desired features/structure).  While it takes a little time, images can have annotation/masks applied prior to processing which should improve output as well.  I hope this helps.

Aaron.

Yes thank you. But I still think the spherical nature works against us.
Just look at the different between datasets. Do you agree?

The Ricoh Theta S website indicates that the images it captures are 5376 x 2688 pixels and says it is equivalent to a 14mp still image and has a 360 degree field of view.  The Samsung Website indicates that the Note 8 has a 12 MP camera with a 77 degree field of view (doesn’t mention resulting image size).  While the Theta can claim a higher MP count, those pixels are spread out over a larger area (360 degree rather than the 77 degrees with the Note 8) and the resulting image has few pixels capturing the area of interest (the Facade/structure).   The benefit of the Spherical image is that you can capture an area with fewer photos, however those photos are at a lower resolution.  Spherical images are helpful with Interior reconstructions where all 360 degrees are capturing relevant areas of interest.  In your example, the area of interest (AOI)  is the building facade and this  more limited AOI doesn’t necessary benefit from the Spherical image.  You could try using a GoPro type camera with Fisheye which would allow you to capture just the facade area with fewer images than with the Note 8 but at a comparable resolution.

Best wishes,

Aaron.

1 Like

Aaron, I have the gopro here today, but it rainy and cloudy. But that is the next test I was going to perform.