Module 5: Supervised/Unsupervised Classification

Comparison of recoded image (supervised classification) with original image of Pensacola, FL. The original image had 15 classes and it was recoded to 7 classes based on the landcover/landtype after supervised classification. All the classes are mentioned in the legend. Supervised classification is based on the knowledge, familarity with the place and pattern recogination skills which results in accurate selection of the training samples. But it was difficult to select the some landtypes especially, grass type which resulted in some errors.

This is an unsupervised classification which is more computer automated where software identifies the statistical patterns in the data without using any ground truth data. Here also, the recoded classes are 7, mentioned in the legend.

Overall, unsuperived classification resulted in better output image than supervised classification. This is because of the reason that it was difficult to identify some classes (grass, forest) in the original image and therefore more errors in the supervised classified image.