Multi-Dimension Feature Learning of Radar Echoes for HFSWR Maritime Surveillance

The dynamically changing ionospheric condition seriously limits the detection performance of High-frequency surface wave radar (HFSWR). In order to identify the covered targets, we separately extract the feature information from phase and time-frequency dimension of different echo components. Then, different dimension features are applied for training the corresponding deep learning networks. When the label results of a radar echo from different trained models are the same, final result of this echo is obtained, in addition, when they are different, this echo information will be added into the training dataset for the next training. Finally, we will find the covered targets under the F-layer specular reflection ionosphere clutter.