Automated segmentation of surface soil moisture from Landsat TM data

Abstract

This study demonstrates a method for satellite remote sensing of surface soil moisture and the automated segmentation of the acquired imagery. The remote sensing method exploits the relationship between surface radiant temperature, vegetation cover, and surface soil moisture. The segmentation process employs a watershed algorithm applied within a morphological image pyramid. This multi-resolution approach compares favorably to fixed-resolution techniques both in computational cost and feature scalability. Applications of both the remote sensing method and image segmentation technique are demonstrated for a Landsat TM image of southwestern Oklahoma.