Improving the accuracy of Multispectral-based benthic habitats mapping using image rotations: the application of Principle Component Analysis and Independent Component Analysis

Abstract

The low number of water penetration bands in multispectral images limits the maximum descriptive resolution and the accuracy of the resulting benthic habitats maps, especially at higher levels of benthic habitats scheme complexities. This research aimed at improving the accuracy of benthic habitats mapping by exploiting the spectral performance of multispectral images using image rotation techniques, which is very beneficial for fast, accurate, and repeatable mapping. Kemujan Island as part of the Karimunjawa archipelago in Indonesia is selected as the study area. Principle Component Analysis (PCA) and Independent Component Analysis (ICA) were applied on Worldview-2 prior to image classification. The inputs for PCA and ICA are deglint bands and water column-corrected bands. Field benthic data collected from photo-transect technique were used to train the rotated datasets in the classification process and to assess the accuracy of the resulting benthic habitat maps. Three levels of benthic habitats classification schemes were constructed based on the variation of benthic habitats insitu, which covers the variations of coral reefs, seagrass, macro algae, and bare substratum. The results show that the application of image rotations on Worldview-2 improves the overall accuracy of benthic habitats mapping and become more effective as the classification scheme complexities increase. In the absence of water column correction, PCA and ICA become the best option to assist benthic habitats mapping.