Automation of MBES noise reduction: An approach based on seafloor bathymetry features derived from manual editing procedures

Abstract

This study introduces an innovative technique specifically designed to effectively mitigate multibeam sonar noise in intricate scenarios where conventional denoising methods have typically faltered. These traditional methods have historically struggled to maintain a delicate balance between sensitivity, specificity, and interpretability, leading many cartographers and research professionals to resort to laborious and time-consuming manual processing techniques. In response to this prevalent challenge, we meticulously formulated a mathematical model to represent each individual step involved in the manual editing process. The approach involves parameterization of cartographers’ perceptual manipulation, resulting in a novel noise rejection methodology. A notable attribute of this method is its simplicity and intuitive nature; users are required to set only a handful of fundamental parameters to obtain highly accurate processing results. Importantly, this technique spares the data from exposure to linear filters such as lowpass or average filters, enhancing the fidelity and reliability of the results. Furthermore, our proposed method retains the interpretability of the processed seafloor bathymetry as it abstains from implementing opaque methodologies such as Neural Networks. This aspect of the technique makes it an ideal choice for researchers and cartographers who prioritize understanding the underlying processes affecting their data. In the discussion segment of this paper, we present a quantitative comparison of the data processed by manual edit, the commercially mature algorithm, and our newly proposed method. This comparison underscores the superior performance and effectiveness of our technique, further reinforcing its potential applicability in real-world scenarios.