Object Oriented Image classification of Marine habitat Using QuickBird multispectral data

Abstract

Remote sensing plays an important role in monitoring the shallow marine habitat of both tropical and temperate environment by spectral response. Classification of marine habitat depends on considering various factors like substrate, wave exposure, salinity and bathymetry at only specific locations. Marine habitats have been at detailed level by means of underwater surveys and biotope mapping in the Sound of Harris, Outer Hebrides, Scotland. In this study, the use of QuickBird imagery and an object based classification has been explored to create more continuous mapping of the broad biotopes that exist in this location. We further explore how the classification may be improved with the help of LiDAR data for providing bathymetric depth data of the area. Results show that whilst the object based approach enables a generalised mapping of some of the main biotopes, many marine biotopes remain spectrally inseparable. Adopting an object based approach could not fully resolve this limitation.