Methodology to distinguish features in complex and congested underwater environments

The underwater community involves researchers who conduct various work and experiments below the surface of Earth’s oceans. Today’s high-frequency sonar limitations do not provide sufficient information for the underwater community. There is a pressing need to increase the accuracy of detection, identify items detected and obtain knowledge of missing details. Currently, no standard methodology exists to evaluate the accuracy and completeness of sonar and detection algorithms; this is critical because both are required to use sonar effectively. Such a methodology is required for the evaluation and comparison of sensors and algorithms. This research investigated a proposed methodology to evaluate effectively underwater perception in complex underwater environments. This methodology considered the challenges of underwater localization and other design factors during the use of a mobile autonomous underwater vehicle and high-frequency sonar, and the employment of underwater signal and image processing techniques. The focus of this research and contribution to the underwater community is an initial robust methodology the underwater community may use to obtain relevant data sets. Furthermore, researchers can apply those data sets to custom perceptors and algorithms, and evaluate whether or not these custom methods are capable of detecting the known features while quantifying the missed features. Ultimately, the results of the signal and image processing, compared to the defined ground truth, demonstrated that the use of the high-frequency sonar was not capable in distinguishing all smaller features and gaps between underwater structures.http://archive.org/details/methodologytodis1094549717Lieutenant, United States NavyApproved for public release; distribution is unlimited