Research On Sea-Surface Target Detection Based On Spatial Correlation Features And Empirical Mode Decomposition

Sea surface target detection,one of the most important research topics in marine science,is a key technology for marine rescue,environmental monitoring,national security monitoring and other applications,and it is particularly significant in the historical context of constructing China into an oceanic great power.In the process of sea surface target detection,the radar echo data includes not only the reflected signals of the target but also the sea clutter,which is the signal scattered back by the complex sea surface.Sea clutter,the main factor affecting the detection results,often masks the target signals due to the complex structure,strong randomness and large fluctuations,which greatly hinders the sea target detection.This paper mainly studies the spatial correlation and empirical mode decomposition of radar detection data and constructs a two-dimensional feature space.Then a detector is designed based on machine learning algorithms to improve target detection performance.The main research contents and results of this paper are as follows.Firstly,this paper studies the spatial correlation characteristics of radar data.When the radar has a higher resolution or works at a low grazing angle,the signal will have strong correlation characteristics in both the time domain and the airspace,but little research has used this feature as a main parameter for detection.In this paper,the classical correlation definition is used to analyze the radar data,and a correlation feature is found through visualization results.This feature can be used to effectively distinguish sea targets,so we treat it as a main parameter for detection.Furthermore,in order to make full use of the random non-stationary information of radar data,a time-variant signal analysis method is introduced to construct a more robust time-variant cross-correlation feature(TVC),thus serving the design of the sea surface target detector.Then,the method based on empirical mode decomposition(EMD)is studied for target detection in this paper.EMD,the most critical step in Hilbert-Huang transform,is an adaptive signal analysis method that focuses on practical physical meaning.For sea surface target detection,EMD decomposition can separate the physical fluctuation trends of different scales in the echo signal,so different characteristics of sea clutter and target signals can be better obtained.By analyzing the components of the radar signal after EMD decomposition,it is found that the sea clutter and the target signal are different in the empirical mode energy distribution,which is extracted as a detection feature,called the relative empirical modal energy(REME).Different from existing methods that only consider single component or low-order components of EMD,this paper combines the highorder and low-order components,which can achieve better detection results.Finally,the time-varying cross-correlation and relative empirical modal energy features are combined to construct a two-dimensional feature space,and the experimental results show that the feature space can distinguish target signal better than single feature.In order to obtain better detection results and achieve a more complete target detection process,a detector is further designed by the SVM algorithm.The experimental results show that the proposed method can effectively improve the detection performance,and the results are more prominent under harsher conditions such as low false alarm probability and low signalto-noise ratio.In the VV polarization mode,the detection probability with a false alarm probability of 0.001 is increased by 10% compared with other methods.