This week, we studied supervised and unsupervised classification and learned about spectral signatures. We learned techniques on how to recognize and eliminate spectral confusion between signatures. Spectral confusion can occur when: the analyst accidentally includes a portion of the another feature in the sample polygon or if the Euclidean distance is set too high. Creating a distance file as reference can help the analyst identify erroneous classifications. Typically, the pixels of the distance file image that are the brightest help identify mistaken classifications.
Here is this week's deliverable map of Germantown, MD:
No comments:
Post a Comment