Mod 5: Unsupervised & Supervised Classification

 This week we learned about unsupervised and supervised classifications. Unsupervised classification and supervised classification are two common methods used to interpret satellite or aerial imagery. Unsupervised is complete without providing training data to classify an image, basically, the system automatically groups pixels into classes. Supervised is a technique where we do provide training data or land cover classes to the system algorithm to guide the classification process. The algorithm uses this labeled data to learn the characteristics of each class (e.i pixels), and then applies that knowledge to classify the rest of the image. It is really neat but does take time to complete. I preferred supervised classification because I was able to train the system to code the image more accurately. We then use these techniques to classify the city of Germantown, Maryland into eight classes. The map below is the product and was created by using the inquire tool, the growing properties to select pixels (verses creating your own polygons), and signature editor features in ERDAS Imagine. The classes chosen were Urban, Road, Deciduous Forest, Mixed Forest, Fallow, Agriculture, Water, and Roads. The area of each in acres is shown next to each class in legend.