Unsupervised segmentation of sand patches from reef images

  • Henry Lee National Institute of Physics, University of the Philippines Diliman
  • Maricor Soriano National Institute of Physics, University of the Philippines Diliman

Abstract

Sand patches, from Philippine reef images, were segmented using unsupervised machine learning techniques. The features used for describing the image patches are a combination of color and texture features. The color features were used for automatically segmenting the target image into clusters of similar color, as defined by the SLIC superpixel algorithm. Each superpixel cluster is assigned a texture descriptor for classification. The texture descriptors used are from the properties of Grey Level Co-occurrence Matrix (GLCM) extracted from pixel values. The clusters were then classified using the k-means clustering algorithm.

Published
2017-06-07
How to Cite
[1]
H. Lee and M. Soriano. Unsupervised segmentation of sand patches from reef images, Proceedings of the Samahang Pisika ng Pilipinas 35, SPP-2017-PB-23 (2017). URL: https://paperview.spp-online.org/proceedings/article/view/77.
Section
Poster Session B (Complex Systems, Simulations, and Theoretical Physics)