Computational coral feature monitoring for the fixed underwater observatory LoVe

Fixed underwater observatories (FUOs) equipped with a variety of sensors including HD cameras, allow long-term monitoring with a high temporal resolution of a limited area of interest. FUOs enable in situ monitoring of visual features like size and color of for instance live cold-water corals using imaging techniques. We present a computational workflow to extract coral features from the huge collection of images recorded by the FUO LoVe (Lofoten - Vesterålen)1. The presented approach allows to represent the image-time-series as numerical values of the features. This enables the images to be subject to statistical analysis to gain insight into the short- and long-term relationships between the corals and the environment. The presented automatic extraction of features from digital photos includes a customized pre-processing of the images (spatial and signal alignment), automated segmentation of the living parts of the corals using an unsupervised learning algorithm and extraction of coral specific numerical features which compensate color shifts due to changes of the in-optical properties affecting the whole image. In this initial study, unexpected temporal change patterns could be revealed, by comparing change patterns of live coral areas with change patterns of coral rubble areas using the CIELab color space.