I had a conflict with a session at SPAR that I really wanted to attend. Dr. Nicolas Vandapel from the Carnegie Mellon Robotics Institute was presenting his research on the use of scene recognition technologies to perform automated 3-D feature extraction. As we all know this is the “Holy Grail” that for the most part has gone unclaimed.
Dr. Vandepel’s position is that the core technology he originally developed to support real-time autonomous ground robot navigation can also be used to automate the process of developing asset inventories for GIS applications from point clouds. His application areas are vegetation management and forest inventory.
If anyone attended this session perhaps you can share some of the highlights. It seems to me this is a promising “out of the box” approach that could hold real promise for certain feature extraction applications, particularly if you limit the number of objects you are searching for to a reasonable number.