Module 5 - Unsupervised and Supervised Image Classification
In this lab, I created a supervised land-use classification of Germantown, Maryland, to visualize current land-cover patterns and support sustainable planning. Maryland’s population has grown 30% over the last 30 years, while land consumption has doubled, making it critical to understand how urban and green spaces have changed.
I used the dataset and selected a band combination that minimized spectral confusion between classes. Using histograms and mean plots, I created and refined spectral signatures for urban areas, roads, water, grass, deciduous and mixed forests, fallow fields, and agriculture. Adjusting parameters like Spectral Euclidean Distance and Neighborhood helped ensure each signature accurately represented its feature.
The final map, recoded into eight classes, includes the classified image, a distance file inset, a legend with colors, class names, and areas, plus essential map elements. This project provided hands-on experience in creating accurate, meaningful land-use maps.

Comments
Post a Comment