Research on Optimal Machine Learning Algorithm for Sea Surface Observation using X-Band Doppler Radar

This study utilizes machine learning algorithms in conjunction with Doppler Radar, proposing to identify the optimal machine learning algorithms and input modes for different sea conditions, and analyzes the impact of wind on observations. An X-Band microwave Pulse Doppler Radar system is used to measure backscattering microwaves at sea surfaces in two irradiation directions. Machine learning models are built to use radar signals which includes Mean Backscattering Power, Mean Doppler Velocity and the Root Mean Square (RMS) of Doppler Velocity to obtain Sea Conditions which are Sea Level, Wind Conditions, Current Conditions and Wave Conditions. This paper proposes the optimal input patterns and machine learning algorithms that can generate the most optimal results concerning sea surface conditions. Furthermore, the study also investigates the impact of wind on radar observations of sea surfaces. Regarding the observing of sea level, Linear Regression is supposed to be the optimal machine learning algorithm. Concerning wind conditions, both Neural Network and Random Forest can be utilized for analysis. However, direct inference of currents through machine learning presents difficulties. Random Forests are the preferred method for wave observation, yielding better results.