Implementation of genetic programming toward the improvement of acoustic classification performance for different seafloor habitats

This is a case study of employing Genetic Programming (GP) on the acoustic backscatter data processing for the classification of different epi-benthos. The purpose is the provision of an improved classification capability that will bring a more reliable understanding of the acoustic backscatter characteristics of different habitats. The result of this study proved that the acoustic classification capability for the recognition of different seafloor habitats can be enhanced by the adoption of GP in the data processing. With a suitable fitness criterion, GP provided an automatic and alternative option to evolve from several initial candidate features into a final compound feature with improved classification performance. Different designs of initial candidate features were also tested to assess the final feature's performance. The execution of GP is illustrated by giving an example of data collected from different seafloor habitats in the Australian coastal waters. The comparison of the classification performance between the present results and those from a previous study without GP is provided. The conclusion is that the implementation of GP in the acoustic data processing can enhance the acoustic classification capability for the characterization of different seafloor conditions.