Aligning points - point limit

[janoz]

I would like to ask why is the point limit by the detection of image specific points predefined to 40 000 and if changing this value has an impact upon the results of the alignment step.

[stihl]

I'm not sure why the default is set to 40 thousand points. I've seen other photogrammetry software that used anywhere between 20k and much much more.

I've tried aligning aerial projects with the maximum number of tie points (setting the value to 0 will do this) and it will find roughly between 200k and 300k tie points per photo. This means that the aliginment takes longer, but I've also noticed that the dense point stage takes far less time. I assume because it has less space to fill in. I've also noticed that the geometry is more complete if the pictures are clear and sharp if I use the maximum tie points allowed.

[Drakhain]

Hi,

I've got a similar question.
I've read on the forum that points below 0.5 reproduction error are recommended and those after .8 are quite inaccurate. So i filtered my point with .5. After that their is not enough points to fully construct the dense cloud.
I do not really understand how to set my alignment parameter.

Can someone please explain how to avoid these disparities in my dense cloud.
Thank you.

[bigben]

I'm not sure why the default is set to 40 thousand points. I've seen other photogrammetry software that used anywhere between 20k and much much more.

I've tried aligning aerial projects with the maximum number of tie points (setting the value to 0 will do this) and it will find roughly between 200k and 300k tie points per photo. This means that the aliginment takes longer, but I've also noticed that the dense point stage takes far less time. I assume because it has less space to fill in. I've also noticed that the geometry is more complete if the pictures are clear and sharp if I use the maximum tie points allowed.

In that situation you'll find that there are problems with the image quality/number of images  or the lens parameters have not adequately been optimised.  You can use higher thresholds, but you will find that the resulting mesh will have more noise in it.  Using lens parameters that you know to be good helps, and after that, you want to look at the images.