Hi All, please find below our new article published in IEEE Access.

R. Deo, J. M. Webster, T. Salles and R. Chandra, “ReefCoreSeg: A Clustering-Based Framework for Multi-Source Data Fusion for Segmentation of Reef Drill Cores,” in IEEE Access, vol. 12, pp. 12164-12180, 2024, doi: 10.1109/ACCESS.2023.3341156.

In this paper, we present a novel framework that utilizes different clustering methods to fuse multiple sources (formats) of data for the segmentation and annotation of reef drill cores. The framework produces a classification scheme for labelling the reef drill core, with segments annotated based on the type of lithologies present in the drill core. We utilize reef cores selected from the Hydrographers Passage in the Central GBR. In our analysis, we combine image data and physical properties data such as bulk density (gamma-ray attenuation), porosity, and electrical resistivity. Then, we evaluate the potential of three key clustering methods including k-means clustering, agglomerative hierarchical clustering and Gaussian mixture models on segmenting the data. 

The primary source of data is based on the image data that embeds information from the color and texture. However, to further distinguish the classes, we also use data from physical properties measurements taken by a multi-sensor core logger on the same core. We follow this by evaluating four selected classification methods for annotating the segmented image that includes support vector machines, multilayer perceptron, random forests, and k-nearest neighbors.

Cheers

Ratneel

#MarineScienceSydneyUni