For this assignment, we practiced using the unsupervised and supervised classification methods in ERDAS IMAGINE to categorize features for land use, land cover purposes. To complete the process, we created a greyscale image showing the distance of difference between color bands, resulting in an image that can be used to assess the accuracy of the supervised classification. In this case, we did the supervised classification on Germantown, Maryland, as seen below:
Although a little confusing at first, I attempted to choose colors that would help viewers make sense of what they were seeing - which is a large amount of agricultural regions, urban development, and a few small waterbodies. We calculated the permeable versus impermeable areas, and created spectral signatures using provided coordinates, seeds, and personal interpretation of water and road features. Tools to find and address any "spectrally confused" areas were used, and then we merged and recoded our classes to simplify our results into eight easily-interpreted categories.
I have continued to experience some quirky glitches in ERDAS IMAGINE, although overall I already feel proficient using the software. In general, it takes awhile for some of the windows to load, including the Metadata and Signature Editor. More than once, I found that more than one window had loaded from clicking too many times.
This assignment was pretty straightforward for me, simply because this process utilized tools similar to what we need at my internship for our QA/QC imagery review. In fact, I am enjoying the land use, land cover classification tasks so much that I am hoping to fit them into my final project for this semester... stay tuned!
Comments
Post a Comment