A Temporal Convolutional Network for Wave Height Estimation from X-band Radar Data

Abstract

A state-of-the-art machine learning based significant wave height (H <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</inf> ) estimation model, which is based on a temporal convolutional network (TCN), is proposed for X-band marine radar in this paper. The input space of the network is composed of three features (i.e., signal-to-noise ratio (SNR)-based, ensemble empirical mode decomposition (EEMD)-based, and gray level co-occurrence matrix based features) extracted from radar images. Two typical H <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</inf> estimation methods (i.e., SNR-based linear regression and EEMD-based linear regression methods) are utilized for comparison with the proposed method using the radar and buoy data collected at the East coast of Canada. It is found that the proposed method can generate the most accurate H <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</inf> results with a root-mean-square error of 0.24 m and a correlation coefficient of 0.94.