Land/water detection and delineation with Landsat data using Matlab/ENVI

摘要

Detecting water bodies and accurately delineating it from the surrounding is an important aspect for various hydrological and agricultural scientific applications. The traditional methods of visiting the actual site and using survey techniques are time consuming and difficult to implement. Remotely sensed data, Landsat in particular, provides a functional means to delineate water boundaries over a large area at a specific point in time. With the multiple bands in Landsat it is important to understand how to use the image and which spectral band and classification method(s) to use for the best hydrological classification. With the result achieved from using Matlab and ENVI software and taking Estuary of Yangtze River in Shanghai, China as case study, it was concluded that band 5 (mid-infrared) in Landsat TM was comparatively best suited for land/water delineation. Furthermore, single-band density slicing with only 2 ranges (water and non-water) appeared to be an excellent method to discriminate water from other land features using Landsat TM band 5.