In this paper we present a novel approach to produce benthic habitat maps from sea floor images in Derwent estuary. We have developed a step-by-step segmentation method to separate sea-grass, sand, and rock from the sea floor image. The sea-grass was separated first using color filtering. The remaining image was classified into rock and sand based on color, texture, and edge features. The features were fed into an ensemble classifier to produce better classification results. The base classifiers in the ensemble were made complementary by changing the weight (i.e. cost of misclassification) of the classes. The habitat maps were produced for three regions in Derwent estuary. Experimental results demonstrate that the proposed method can indentify different objects and produce habitat maps from the sea-floor images with very high accuracy. © 2013 Copyright the authors.
CITATION STYLE
Rahman, A. (2013). Benthic Habitat Mapping from Seabed Images using Ensemble of Color, Texture, and Edge Features. International Journal of Computational Intelligence Systems, 6(6), 1072–1081. https://doi.org/10.1080/18756891.2013.816055
Mendeley helps you to discover research relevant for your work.