kNN for image classification

In this video I explain how kNN (k Nearest Neighbors) algorithm works for image classification. We vary the parameter max distance of neighbors to be classified (from 1 to 100), in order to show the evolution of the classification. I selected 9 samples for 3 patterns (bare soil, urban areas, vegetation) and used k=3. I provide one animation of the scatterplot in 3 visible spectral bands of a CBERS satellite image (available at github.com/tko....
The source code used to create the animations is available at github.com/tko...
The slides are available at prezi.com/bf7r...
Download free remote sensing images at www.dgi.inpe.br...
Please like and share the video, and subscribe to my channel.