Module 5: Supervised/Unsupervised Classification

    For this week's lab, I used different classification methods to classify land use/land cover of different areas. Throughout the lab, I explored ways to gather information that feeds into the classification (supervised) process and ways to avoid spectral confusion during the classification process. The selection of the bands used in the classification plays a big role in the quality of the output. Last, I learned how to interpret some of the output generated during the classification process such as the distance file.

    For exercise 5, I was given a raster of Germantown, MD, and I was asked to classify the image using a supervised classification. Initially, I had difficulty properly classifying the roads. Personally, I think an image with a higher spatial resolution would have been better for classifying the roads, but after removing some of my initial road signatures that I made and adding more agriculture and deciduous forest signatures, I was classifying the roads better. The final product is not perfect. Some of the agricultural areas could probably be better classified, but this output is much better than the initial output that I had. The distance file indicates that most uncertainty in the classification is around some of the rural areas. Below is an image of my classification output for Germantown, MD.