Top-down analysis

Examples of high-resolution towed sidescan sonar data of cold-water coral reefs, acquired on successive passages (top), and the results of automated image classification using Grey Level Co-occurrence Matrices (bottom). Note the different outcomes caused by the different sonar/seabed configuration.

Robust mapping of habitats and marine landscapes on the regional to medium scale

Work package WP1 involves:

  • Development of objective classification methods for deep marine landscapes
  • Evaluation of different automated image segmentation techniques for (AUV/ROV-based) multibeam and sidescan sonar data
  • Performance assessment through replicate surveys
  • Development of a confidence assessment methodology

 

Results:

  • Objective automated classification technique for marine landscape mapping in complex deep-sea environment, including confidence assessment
  • New toolbox for the ArcGIS 10.x software to support automated image segmentation. Also summarised in a 2-page factsheet