Hello,
I am a complete beginner in photogrammetry. I’d like to create a model of linear object but RC seems to have some issues with creating point cloud. I’d say it’s because incorrectly shooted photos. Should I stick to specific method of shooting? Can you give me some advice/tips to shoot it? I went through the basic tutorials, but it’s aimed more on common object than linear/long. Basics of shooting will be the same I guess, but wanted to ask if there are some rules to keep when shooting objects like that.
Thanks for replies
Shoot using a drone, use a flight pattern, fly at different heights, multiple angles (to cover areas underneath trees for example), …
And overlap, overlap, overlap.
You can never shoot too many photos 
Did I mention overlap?
If you have access to accurate GPS make use of it.
And make sure you have enough overlap
ShadowTail wrote:
Shoot using a drone, use a flight pattern, fly at different heights, multiple angles (to cover areas underneath trees for example), …
And overlap, overlap, overlap.
You can never shoot too many photos
Did I mention overlap?
If you have access to accurate GPS make use of it.
And make sure you have enough overlap
Thank you for reply Shadowtail.
Yeah that’s exactly what I though, but hoped it’ll be possible without drone. Is there a chance to get accurate result without it? Anyways, if we talked in number (I know it’s really hard to say, because it depends on many factors, plus the more photos the better), but what’s the minimum for let’s say 20m segment, 4m wide river? Again, I know it really depends… But want to get at least rough idea.
For something that “small” you can likely get away without using a drone.
The number of photos required depends on the location though.
My best guess would be somewhere between 500 and 1000 photos if you want to capture every little detail.
Less if you don’t need sub-millimeter accuracy.
Either way, the water is going to cause you issues. So you’d likely have to work with control points.
If you can return to the capture area you can shoot more photos as needed if what you got the first time around isn’t enough.
Mind, though, that I am basing this on the camera I use, which only takes 12MP photos at best.
If yours does better you can get away with taking fewer photos.
But maybe someone else has more/better suggestions
Ok. So I’ll focus on overlaping. The river was only for example, actually want to shoot road excavation, but the shape is quite similar to river basin and easy to explain and understand. Thank you again for valuable advice. Really appreciated
curb.jpg
I took 32 photos using my mobile phone to create this little patch of scenery.
The result, especially the texture, would have been noticeably better had I used a decent camera and a tripod.
But for hand-held pictures from a phone this is about as good as it gets.
When you are in a hurry, that is
Also, this is a preview quality render using my own fine-tuned settings. Its somewhere inbetween the default preview and normal quality reconstruction.
Hey ST,
do you mind sharing your settings?
And how about a wireframe? :ugeek:
And for the main question:
It is also important to get as near to vertical (to the surface you are interested in) as possible.
Steep angles will not produce good results.
Also, you need to be careful about slowly changing angles, if that is the case.
Oh, and has overlap been mentioned yet? :lol:
Wow, important note with vertical angles. Didn’t know it’s that different in comparison with more like horizontal.
Almost forget about the overlap, thank for reminding 
Pretty much realized that it’s better to shoot as many images in as big resolution as possible, than to play around with more advanced settings. Right?
Yes, the stress being on “playing around”.
It takes time to get better results.
What I also underestimated a bit is in-camera undistortion of JPGs.
While that is no problem with older cameras, the new ones are prone to have it (especially at the lower end of the price spectrum)…
And oh yes, the more horizontal, the badder it is (to quote Black Books :D).