3dpalign

By Dan Azuma and Brad Nelson


Getting started...

You will be able to use 3dpalign to take two or more range scans and find a very accurate common coordinate frame for the scans. To do this you will need to provide a few things to the program. You will need a set of range images each with the .range extension, also you will need corresponding .params files (see this page for a detailed description of the file format). After preparing these files you will need to list the name of each one, without the extension, in a .3dp file. For example, if you gave acquired 4 views of your favorite statue, you can generate a file statue.3dp containing all the views:

statue1
statue2
statue3
statue4

You may then run 3dpalign on that file. It is located on ugcs in ~pinky/bin/3dpalign. If you are interested there is a windows95 version. If you have an opengl driver for your system you can give this one a go. It is located in ~pinky/bin/3dpalign.exe

3dpalign takes one argument, the name of a .3dp file. Run for example:

3dpalign statue.3dp

to align the 4 views of the statue.

Once the program has load you should see 4 windows. The two colored windows at the top are for viewing pairs side by side and selecting point correspondences. The gray windows on the left will show progress you have made with aligning the current view. The lower right hand window is for selecting which views to align.

In each window there are menu options that can be selected by holding down the right mouse button. In the pair viewing windows this will allow you to select if you are adding points, checking point pairs, or deleting points. In the selection window this will allow you to select a view as the red or blue view.

In each of the pair views and in the aligned view window, the left mouse button is used to rotate the viewpoint. Holding down ctrl allows scaling.

In the pair view windows points are selected or added by use of the middle mouse button, or if only two button are available with shift and the left mouse button.

Alignment...

The process of alignment consists of making point correspondences between each view by adding point first to one view, then to the other. The red point in a view is the currently selected point, the blue points are set points. After this has been done for a connected set of the views. A fast global align should be done in the alignment window. If this is an adequate start for the full solver, then a global alignment step should be done. If not, more point correspondences should be added.

If you have any questions about the interface contact: bdnelson@cco.caltech.edu

* RETURN TO PREVIOUS PAGE