Quantitative seafloor characterization using angular backscatter data of the multi-beam echo-sounding system - Use of models and model free techniques
Abstract
For quantitative seafloor roughness characterization and classification using multi-beam processed backscatter data, a good correlation is indicated among the power law parameters (composite roughness model) and hybrid ANN architecture results. Moderate classification efficiencies are achieved in the range of 32% to 46% using the SOFM classifier. Implementation of a neural-based hybrid classifier uses SOFM as a precursor to broadly identify the number of classes in the input space. For further improved classification exercise of multi-beam backscatter data, supervised learning algorithms of LVQ1 and LVQ2 have been successfully adopted. Present study emphasizes that unprocessed (raw) backscatter data provides successful real-time classification of seafloor roughness using hybrid ANN.
Collections
- Scholarly literature [8950]