A Hybrid Vision-Based Method of Encountered Wave Field Measurement for Navigating Surface Vehicles

Ocean waves have a significant impact on surface vehicles, such as ships and unmanned surface vehicles (USVs), and measuring wave fields using existing methods is inefficient and costly. To address this issue, this article presents a hybrid vision-based approach for navigating surface vehicle using monocular image of wave field and vessel motion data. The proposed method is based on depth estimation of image pixels and considers the effect of vessel motion on detection accuracy. The depth of each image pixel is equalized to the sum of the depth of the mean water surface and the depth variation induced by wave elevation. The depth of the mean water surface can be determined using vessel motion, and the depth variation can be approximated by a learning model using monocular images. The learning model was trained using images of field measurements performed in the South China Sea near a typhoon, and
${X}$
-band radar data were acquired for comparison. Using the proposed method, a 3-D wave field can be reconstructed. Wave spectra and wave parameters such as the dominant frequency and significant wave height were analyzed to validate measurement accuracy. The expected errors were also analyzed. The performances of the proposed and other vision-based methods were evaluated using the same dataset and compared in terms of accuracy. The results demonstrate that the proposed method exhibits an improved performance in reconstructing the wave field elevation.