Research On Marine Surveillance Radar Clutter Suppression And Small Target Detection Methods Based On Machine Learning

As an important part of marine environment perception and coastal defense system,marine surveillance radar has important application value in marine environment monitoring and other fields.In reality,the target detection performance of marine surveillance radar will be seriously affected by complex marine environment and other factors.Therefore,effective sea clutter suppression and high-precision sea surface target detection technologies are of great significance for improving marine surveillance radar’s detection performance.Among which,sea clutter suppression and sea surface small target detection based on machine learning have been widely concerned.Aiming at the high performance detection of marine surveillance radar sea-surface targets,this thesis focuses on sea clutter modeling and analysis,sea clutter suppression and sea surface small target detection based on machine learning.The specific contents are shown as follows:1.The scattering characteristics and spatio-temporal correlation characteristics of sea clutter are analyzed,and the classical sea clutter amplitude distribution model is studied.Based on the correlation composite K distribution model,the simulation and generation of sea clutter data in the scanning mode of marine surveillance radar is completed,which lays a data foundation for the research of sea clutter suppression and target detection methods.2.A clutter suppression algorithm based on Clutter Cancellation Generative Adversarial Network(CCGAN)is proposed.A sea clutter suppression model is established,and a clutter cancellation structure based on skip connection is studied.The sea clutter characterization network is designed to realize the effective learning of sea clutter.In addition,by introducing the target information preserving loss,the target information is protected while the sea clutter is effectively suppressed.3.A small target detection method based on Feature Parallel Extraction Target Detection Network(FPE-TDN)is proposed.A small target detection model is established,and the object feature extraction structure based on parallel convolution is studied.A small target detection network is designed,the effective detection and extraction of seasurface small targets are realized.This method improves the detection rate of small targets,reduces the false alarm rate effectively,and realizes the high performance detection of sea surface surveillance radar targets.The experiments results based on simulation and measured data show that CCGAN can effectively suppress sea clutter and protect sea surface target information,and FPETDN can accurately detect sea-surface small targets.