Home       Chase County Overview       Makesig       Maxlike       Mindist       Piped       Cluster   Isocluster       Results       References

  


Cluster


Cluster is an unsupervised classification technique. It is based on signature values from a specified composite image.

Before running cluster, I used convert to convert my 234 composite from integer binary to byte binary. Running the cluster module I chose to analyze the bite binary 234 composite. I did a broad generalization, and set the maximum number of clusters to 15. I studied this 15-cluster image to determine which of the clusters were associated with crops, forest, grassland, pavement, soil, and water. I then used reclass to group the 15 clusters into 6 clusters which were associated with the crops, forest, grassland, pavement, soil, and water classes. I then created this image with the six clusters appropriately labeled as the features they represented.


The cluster classification technique did a poor job of separating and correctly classifying many features. The cluster image shows riparian vegetation as crops, many bare soil areas as water, and some bare soil areas as rock/pavement.


This pie graph displays a statistical summary of the cluster classification results.




  
Home       Chase County Overview       Makesig       Maxlike       Mindist       Piped       Cluster   Isocluster       Results       References