Small Target Detection Method on Sea Surface Based on T-GCN

• Published in: 2023 IEEE 16th International Conference on Electronic Measurement & Instruments (ICEMI) pp. 88 - 92
• Main Authors: Zhou, Xing, Xing, Hongyan, Ye, Ru
• Format: Conference Proceeding
• Language: English
• Published: IEEE 09.08.2023
• Subjects: Correlation; Logic gates; Object detection; Predictive models; Radar clutter; recurrence plot; sea clutter; Sea surface; small target detection; temporal graph convolutional network; Time series analysis
• DOI: 10.1109/ICEMI59194.2023.10270595

Improving the detection ability of small targets in sea clutter is always a popular subject in the field of radar technology. The traditional sea clutter prediction method only considers the fluid alteration of radar echo, but neglects the spatial correlation. This research proposes a new temporal graph convolutional network (T-GCN) detection technique for tiny sea surface targets. Multiple historical sea clutter sequences are input by recurrence plot (RP). In this case, the graph convolutional neural network extract spatial characteristics from numerous sequences by capturing their topological structure. The time series is input to the gated recursive unit. Prediction result is output using the time characteristics. Experiments show that T-GCN can get spatiotemporal properties from sea clutter model, and the prediction results are better than that of support vector regression model, temporal dynamic model constructed by gated recursive unit and spatial correlation model constructed by graph convolutional network, which increases the visibility of small sea surface targets and offers a fresh concept and method for predicting sea clutter.