More of Images

Done with the first day. Looking at the documentation I can tell you who is involved with this class. Robin Weeks is the Research assistant Professor of the Department of Urban Design and Planning at the Univ of Washington. He has told us he moved on (private world), but is still doing this class. Duncan Munro is an IT professional of Watershed Management division of Seattle Public Utilities. They are the teachers. They are also both British. They are image scientists, though, and are teaching this like a crash course in remote sensing. It's really good so far.
That's not to say that there haven't been sleepy moments. There was an hour where Duncan was droning on about project design and planning. There was some big spreadsheet he made in excell. That's all I remember.
Other than that we have been playing with ENVI. It's a pretty fun program, and very fast, although I don't know how big the images we are using are compared to what you would be using in actual practice.
We did exercises with some multispectral Landsat data. We had 6 bands (three bands of IR) and used several analysis tools to compare the bands and find out which ones (and which combinations of bands) identified certain types of vegitation, or lack of. We also discussed some of the limitations.
One of the limitations was that the automated classifications didn't do a great job of figuring out patterns in the images. They did alright, but missed variations caused by shading in high topography areas. They also clumped too much. We ended up doing a better job doing it manually (supervised classification) using scattergrams.
It's really neet seeing how land patterns look when looking at the whole spectra. For instance, heavy vegitation has a curve that rises in the green spectrum, falls again and then rises in the near infrared before falling. ENVI allows you to see a full spectrum curve for each pixle you select. It's really interesting.