3d model/point cloud from photos - manually

Wondering if anybody has come across software that can do this.

I have found several that attempt to perform the matching of photos to create a 3d point cloud (e.g. photosynth, Insight3d, VisualSFM) which use automatic feature recognition (RANSAC) - however these are very hit and miss.

What i’m looking for is a program that will let me assign the points to features on the photographs manually building up my own point cloud as I go. Obviously this will be much slower than the automated methods - but hopefully it’ll be much more accurate.

The program Hugin allows you to assign points to features manually - but it’s designed for stitching panoramas rather than generating point clouds. I guess what i’m after is a 3d version of Hugin.

Any ideas?

edit: Found this video online. It looks like software designed for Forensics - but seems to do exactly the type of thing i’m looking for.

I have great interest in this area as well.
I want to develop some method for it in blender but have not had the time yet.

The only one I know of is Photomodeler.


Moved from “General Forums > Blender and CG Discussions” to “Support > Other Software”

I’ve not gotten a chance to do it yet but what you might be able to do is take a structured light photo. I used a transparency with lines printed on it to get this.

And then take a mesh and move it to match the lines in only the one axis direction, and then select each line as an edge loop and rotate them to form a topology. Of coarse you would have to mirror it and then sculpt to get a better model etc.

Yeh - had a look at that - quite pricey though.

I guess Blender could kinda do it if you could set background images based on arbitrary camera views.

You could set the camera in one position, set the background image - then create two or three points.

You could then create a second camera position with another image - then move the camera so that the points line up with the new background image.

Repeat as necessary.

I guess the camera tracking function does kinda already include this type of thing - I wonder would it be a big job to make it work across multiple images taken from different angles (I guess that is technically what it does anyway).