Recognition and assessment of seafloor vegetation using a single beam echosounder

Abstract

This study focuses on the potential of using a single beam echosounder as a tool for recognition and assessment of seafloor vegetation. Seafloor vegetation is plant benthos and occupies a large portion of the shallow coastal bottoms. It plays a key role in maintaining the ecological balance by influencing the marine and terrestrial worlds through interactions with its surrounding environment. Understanding of its existence on the seafloor is essential for environmental managers.Due to the important role of seafloor vegetation to the environment, a detailed investigation of acoustic methods that can provide effective recognition and assessment of the seafloor vegetation by using available sonar systems is necessary. One of the frequently adopted approaches to the understanding of ocean environment is through the mapping of the seafloor. Available acoustic techniques vary in kinds and are used for different purposes. Because of the wide scope of available techniques and methods which can be employed in the field, this study has limited itself to sonar techniques of normal incidence configuration relative to seafloors in selected regions and for particular marine habitats. For this study, a single beam echosounder operating at two frequencies was employed. Integrated with the echosounder was a synchronized optical system. The synchronization mechanism between the acoustic and optical systems provided capabilities to have very accurate groundtruth recordings for the acoustic data, which were then utilized as a supervised training data set for the recognition of seafloor vegetation.In this study, results acquired and conclusions made were all based on the comparison against the photographic recordings. The conclusion drawn from this investigation is only as accurate as within the selected habitat types and within very shallow water regions.In order to complete this study, detailed studies of literature and deliberately designed field experiments were carried out. Acoustic data classified with the help of the synchronized optical system were investigated by several methods. Conventional methods such as statistics and multivariate analyses were examined. Conventional methods for the recognition of the collected data gave some useful results but were found to have limited capabilities. When seeking for more robust methods, an alternative approach, Genetic Programming (GP), was tested on the same data set for comparison. Ultimately, the investigation aims to understand potential methods which can be effective in differentiating the acoustic backscatter signals of the habitats observed and subsequently distinguishing between the habitats involved in this study.