Sequence-Feature Detection of Small Targets in Sea Clutter Based on Bi-LSTM
Hao Wan, Xiaoqing Tian, Jing Liang, Xiaofeng Shen
Abstract
In the complex and changeable marine environment, it is difficult to achieve high-performance detection based on statistical theory for small sea surface detectors in practical applications. In this work, a method for detecting small targets on the sea surface based on sequence-features is proposed from the perspectives of feature extraction and feature classification on a basis of real data. The sequence-feature method firstly extracts three sequence-features from the radar echo sequence, namely, the instantaneous phase, Doppler spectrum entropy, and short-time Fourier transform (STFT) marginal spectrum. The features are used to train the bidirectional long short-term memory networks (Bi-LSTM) model so that it has the ability to distinguish the target and the sea clutter. The measured dataset verifies that the proposed detector can achieve the detection of sea surface targets with controllable false alarm rate, and the performance is better than several existing detectors.