The above satellite image was classified and then recoded using ERDAS 9.3. The ISODATA algorithm was used with 15 classes to be identified. From this the recode process narrowed the number of classes to 8 by combining similar classes.
Although the image roughly portrays the land characteristics on the real ground there are issues. Agricultural areas had roughly the same color as the airport runways so both were identified in the same class. Also, trees look the same whether in an urban or forest area and tree canopies tend to overwhelm parts of the urban area especially in a geographic area with prolific plant growth such as Pensacola, Florida.
These are typical problems in an unsupervised classification. One potential fix would have been to use more classes and a better color scheme to identify the different land use areas. This is primarily true for the vegetation aqnd urban areas that were difficult to differentiate from one another.
No comments:
Post a Comment