SALA-LSTM: a novel high-precision maritime radar target detection method based on deep learning

Area of Science:

• Marine surveillance
• Radar engineering
• Artificial intelligence

Background:

• Pulse-compression radar is crucial for low-cost civil maritime detection.
• Complex sea clutter significantly degrades radar echo quality and target detection.
• Existing mathematical methods struggle to differentiate sea clutter from maritime targets.

Purpose of the Study:

• To develop an advanced deep learning-based method for maritime radar target detection in sea clutter.
• To improve the accuracy and reliability of detecting maritime targets amidst environmental noise.

Main Methods:

• Proposed a novel Self-Adaption Local Augmented Long Short-Term Memory (SALA-LSTM) structure.
• Integrated adaptive convolution into LSTM cells to enhance local correlation perception.
• Developed a deep neural network incorporating SALA-LSTM for radar target detection.

Main Results:

• The proposed network demonstrated superior detection performance.
• Evaluated using a measured dataset with diverse maritime scenarios.
• Achieved improved detection probability and reduced false alarm rates compared to existing methods.

Conclusions:

• The deep learning-based SALA-LSTM network offers a significant advancement in maritime radar target detection.
• The method effectively handles complex sea clutter, improving marine environment monitoring.
• This approach provides a more robust solution for identifying maritime targets.