Indexing in large CAD/Shape Databases

Given large databases of shape and object models, possibly distributed over a LAN or WAN, the task is to retrieve pieces in a catalogue of parts, given some similarity measures in terms of shape or image features.

Imagine the engineer is looking for existing shape models of certain parts of a mechanical design, from different Ford or Mercury design offices. A fast query is required. The user may want to:

1.

use an image of the part,

2.

draw a sketch of the part,

3.

specify a known CAD model, or subset thereof (in terms of certain features).

For each case, different data mining techniques will be required. However, with out common shape architecture such tasks become reasonable.