Improving bathymetric images exploration: A data mining approach

Bathymetry is the science of measuring and charting the depths to determine the topography of the seafloor and other bodies of water. It has several important practical and academic applications. For this reason, having computational tools capable of analyzing bathymetric charts would be useful for domain experts studying the various problems related to water depth. Data mining is a well known technique for extracting information from large datasets, but cannot be directly applied to images. The contribution of this work is an approach for using data mining in bathymetry images. We propose a method for processing input images, in order to extract records and their features, which can be processed by classic data mining algorithms. Additionally, we also propose techniques to visualize both data mining results and map characteristics. For evaluation purposes, the proposed approach was applied to a cold-water corals dataset, in order to predict where corals are likely to be found, under a domain expert supervision.