In Module 10 we learned to create a supervised classification of data. This process begins with manually creating signature points. These points train the software to look for similar pixels and assign them to an appropriate category. Once all categories are created, and each category may have multiple signature points, it is wise to look for Spectral Confusion.
By looking at Histogram Plots and Mean Plots we can see which categories may be overlapping, thus assigning pixels to the incorrect classification. Creating Distance image helps to identify areas that are likely to be mislabeled. After using above tools to determine which spectral bands are closest together, we can change the spectral bands of the image to minimize Spectral Confusion. Once all that is done we can combine like categories into a single category, such as Agricultural 1 and Agricultural 2 into Agricultural.
Once above steps are done we move from ERDAS Imagine to ArcMap to complete the map.
The map on the left shows unsupervised classification of an area.Unsupervised classification relies on the software to classify the pixels. Once classified I manually assigned each type of pixel into one of five categories, and gave each category a unique and distinctive color.
Most of the process was done in ERDAS Imagine. First, to classify the image into 50 distinct pixel categories. Then, to manually assign each pixel to a new category. Lastly, ArcMap was used to create the final map.
