In this homework you will take a complete set of views of your object and align them using 3dpalign. 3dpalign should be installed on UGCS. If you experience any problems e-mail the TAs. Read the tutorial before starting.

For each view you will need the ".params" file and the ".range" file. Please refer to the 3dpalign tutorial for more information.

It is extremely important to get accurate calibration results. If you have consistent errors in the reconstruction you will find it impossible to align the views.

It's a good idea to scan in your object with a plane behind it (to help you in the next homework with the volume carving). If you do this, make 2 .range files for each view - in one of the files, remove the points corresponding to the plane. You will do the alignment on the ".range" files that do not contain the plane, and then use the updated ".params" files with the ".range" files that do contain the plane for volume carving.

Also, it is a good idea to clean your data. You would have probably seen in before that there are lots of noisy points, "floating" around your object. Creating the mesh got rid of most of these, since they weren't close to anything. However, since you are not using the mesh for alignment, but the point cloud, these random points might sabotage the process. If you haven't already done so, you should write a MATLAB function that displays the point cloud from several viewing angles allowing you to remove points from the cloud.

Add to your results web page several screen captures of the aligned cloud of points (use "snapshot" on the SGI's). This should be done after multiple iterations of the global alignment function in 3dpalign.

Also add a thumbnail picture for each of the views, pointers to the updated (after alignment) ".params" and ".range" files (each of the ".params" files should contain a 4x4 alignment matrix). Make sure you understand the format of this rigid body transformation matrix.