- Pattern recognition is one of the session topics at Laser Scanning ’09.
- There is not a lot of published info on the use of pattern recognition techniques in remote sensing.
- With point cloud densities increasing it would seem that pattern recognition is an important technique for automating the feature extraction process.
I noticed in the preliminary program for the ISPSRS Laser Scanning ’09 workshop that there is going to be a session on pattern recognition. I don’t know a lot about this approach, and from the research that I attempted to do it appears that it is discussed more with respect to machine learning than remote sensing.
The titles of a few of the papers are revealing – “Detection of Curbstones in Airborne LiDAR Data”, and “Automatic 3D Powerline Reconstruction Using Airborne LiDAR Data”. These would appear to be focused on identifying linear objects, although there was also a paper on identifying tree crowns.
The powerlines would certainly seem to be a likely candidate considering the background where they are often found, and for that matter tree crown shapes, which used to be used to manually identify tree species in the old photo interpretation days, are also well defined patterns that could be a good match for working with 3D point clouds.
From my limited research I did find that there is the International Association for Pattern Recognition. I suppose that is not a surprise. A quick look at their website does not reveal a strong link with remote sensing.
The ASPRS website search did turn up a number of references, but nothing that linked laser scanning feature extraction with pattern recognition. With point cloud densities increasing it would seem like this is fertile ground for automating the feature extraction process.
If you have any info on this, please let me know.