Unsupervised Partitioning of Field Reflectance Data with K_means

[braimahm]

The code constructs K_means clustering algorithm to processes the field reflectance data. The code interpolates the 1nm reflectance to 10nm computes derivatives, and construct k_means model for field data. The objective of the project is to partition agricultural field reflectance data. K-means model was constructed to partition agricultural field cover after harvesting. The data was first transformed to derivate spectra to bring out unique features. The K mean model generated 3 classes of agricultural field cover classes. The clusters are very great because it shows few overlaps in the dataset. The first one is a mixture of corn residue and soil; the second class is mainly about soil with few corns' residue. The third class shows weeds and grass. The kmeans model does a good job but it will be great if it can differentiate weeds from grass and dried corn residue from bare soil.

Unsupervised_Machine_Learning_Model_Colaboratory.pdf