Posted on June 21, 2013 by Scott Honda
Today I processed the preliminary data. After a few attempts, I finally managed to make a classification of the reef area around Kualoa Ranch/Park and Ka’a’awa valley in Erdas. First I masked out the land so only the offshore area was in the image. Then I did a principle component analysis of that masked image. I took the first 4 bands as the standard diviation of the 4th band was around 20 and I felt anything below that would be useless. I ought to make some extra classifications using other combinations of bands 1 – 4. K-means classification was then used on the image over 15 iterations with 36 possible classes. I think it worked well. I ended up making 7 overall classes, Sand, Reef, Open Ocean, More Sand (than reef), More Reef (than sand), waves, and unclassified. It’d be good to do some sort of ground truthing on my classification, but I am not sure if I’d have time to do it, not to mention that I would need to take a Trimble out into the ocean. But overall I believe the classification matches pretty well after overlaying it on the original imagery.
After I finally managed to make my classifcation, I tried to do some interpolation of the the off shore LiDAR data to create a bathymetry layer. Arc must have failed to execute about 20 times! But now I think I fed Arc everything that it need to be happy and process my data! Hope that turns out well. Once this is done I want to try and combine the classification with the bathymetry to try to pin point archaeological artifacts and/or reef fish habitats.
Anyway, once this is done I’ll be ready to get back into the field and gather some data.