Tuesday, 13 August 2002
Statistical Clustering for Hierarchical Storm Identification
We describe a recent method of
partitioning the pixels of an image such that the partitions
at one step are wholly nested inside the partitions of the next
step. This is achieved through K-Means clustering followed by
region growing and morphological processing.
By steadily relaxing the inter-cluster distance between the clusters
that is allowed by the morphological processing, a hierarchical
tree of clusters is obtained.
We show in this paper that this hierarchical tree of clusters can be used to identify storms at different scales on Weather Service Radar (WSR-88D), Terminal Doppler Weather Radar (TDWR) and on GOES satellite images.
Supplementary URL: