Evaluation and improvement of methods for estimating sea surface wave parameters from X-band marine radar data

Yang, Zhiding (2022) Evaluation and improvement of methods for estimating sea surface wave parameters from X-band marine radar data. Masters thesis, Memorial University of Newfounadland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (6MB)

Abstract

In this thesis, several algorithms have been proposed for estimating ocean wave parameters from X-band marine radar data, i.e., wave direction, wave period, and significant wave height. In the first part of this study, the accuracy of wave direction and period estimation from X-band marine radar images under different rain rates is analyzed, and a sub-image selection scheme is proposed to mitigate the rain effect. Firstly, each radar image is divided into multiple sub-images, and the sub-images with relatively clear wave signatures are identified based on a random-forest based classiffication model. Then, wave direction is estimated by performing a Radon transform (RT) on each valid sub-image. As for wave period estimation, a random-forest based regression method is proposed. Texture features are first extracted from each pixel of the selected sub-image using the gray-level co-occurrence matrix (GLCM) and combined as a feature vector. Those feature vectors extracted from both rain-free and rain-contaminated training samples are then used to train a random-forest based wave period regression model. Shore-based X-band marine radar images, simultaneous rain rate data, as well as buoy-measured wave data collected on the West Coast of the United States are used to analyze the rain effect on wave parameter estimation accuracy and to validate the proposed method. Experimental results show that the proposed subimage selection scheme improves the estimation accuracy of both wave direction and wave period under different rain rates, with reductions of root-mean-square errors (RMSEs) by 6.9ﾟ, 6.0ﾟ, 4.9ﾟ, and 1.0ﾟ for wave direction under rainless, light rain, moderate rain, and heavy rain conditions, respectively. As for wave period estimation, the RMSEs decrease by 0.13 s, 0.20 s, 0.30 s, and 0.20 s under those four rainfall intensity levels, respectively. The second part of research focuses on the estimation of significant wave height (Hₛ). A temporal convolutional network (TCN)-based model is proposed to retrieve Hₛ from X-band marine radar image sequences. Three types of features are first extracted from radar image sequences based on signal to noise ratio (SNR), ensemble empirical mode decomposition (EEMD), and GLCM methods, respectively. Then, feature vectors are input into the proposed TCN-based regression model to produce Hₛ estimation. Radar data are collected from a moving vessel at the East Coast of Canada, as well as simultaneously collected wave data from several wave buoys deployed nearby are used for model training and testing. After averaging, experimental results show that the TCN-based model further improves the Hₛ estimation accuracy, with reductions of RMSEs by 0.33 m and 0.10 m, respectively, compared to the SNR-based and the EEMD-based linear fitting methods. It has also been found that with the same feature extraction scheme, TCN outperforms other machine-learning based algorithms including support vector regression (SVR) and the convolutional gated recurrent unit (CGRU) network.

Actions (login required)

View Item