Notes: This is a subset of a larger Bay Area map that was extracted through the use of an Inquire Box.
NDVI Transformation
Notes: The Normalized Difference Vegetation Index makes use of the inverse relationship between red and near-infrared reflectance associated with healthy green vegetation. One disadvantage is its sensitivity to canopy background variations.
Tasseled Cap Transformation
Notes: An important vegetation index that can break down the imagery into brightness, greenness and wetness variables. It is very useful in detecting soil moisture.
Tasseled Cap Transformation
All Band 1 (Brightness)
Tasseled Cap Transformation
All Band 2 (Greenness)
Tasseled Cap Transformation
- All Band 3 (Wetness)
Principal Components Analysis (PCA)
This is a technique that transforms the original remotely dataset into easier to interpret set of uncorrelated variables that represent most of the information from the original dataset. Above is the original image comprising of 6 components. Most of the variance can be accounted for within the first three principal components as will be shown in the pictures below. Components beyond the first three will account for less variance and could account for some noise.
1st Principal Component (All Band 1) 2nd Principal Component (All Band 2)
3rd Principal Component (All Band 3) 6th Principal Component (All Band 6)
Unsupervised Classification
This classification method requires only minimal amount of initial input, most importantly the number of clusters.
In the following pictures, one image of Redwood Shores was taken then classified into 8 clusters. In the other image, the entire Bay Area was classified into 8 clusters, and then the Redwood Shores subset was cut out for comparison. Based on the original image, the clusters will be classified differently as can be seen in some slight differences in colors for certain areas.
Redwood Shores Image classified into 8 clusters.
Entire Bay Area (only Redwood Shores subset shown) classified into 8 clusters.
Supervised Classification
This method require much more initial input as training data must be provided for each class that the user requires. Two of the most popular supervised classification methods are maximum likelihood and minimum distance. The images below show how the classifications will differ between methods although the same number of classes (5) and training data were used