Research On Small Target Detection Method On Sea Surface Based On Multi-domain And Multi-dimensional Feature

With the development of stealth and miniaturization of targets,the detection of small targets on the sea surface has become an important and difficult task for marine radar to monitor.Because the scattering cross section of small targets is small and the moving speed is slow,the target echo is submerged by the strong sea clutter in the time domain and covered by the broad spectrum of the sea clutter in the frequency domain,which greatly increases the detection difficulty.Aiming at the problems of low detection probability and high false alarm rate of traditional detectors,this paper mainly studies the small target detection method based on multidomain and multi-dimensional features.By analyzing the data of multiple domains,the refined differences of sea clutter and echo with targets are deeply explored,and these differences are condensed into multiple features.Then,multiple features are combined for detection to further improve the radar’s sea detection performance.The main research content of this paper is as follows:Firstly,the basic theory of target detection in sea clutter is introduced.First,the multidomain characteristics of sea clutter are analyzed from the aspects of statistical distribution and time-frequency characteristics.Then,the basic problem of sea surface target detection is described by the binary hypothesis test.Finally,the feature detection methods are introduced,and the multi-dimensional feature extraction methods are listed,which lays a theoretical foundation for the subsequent research of multi-dimensional feature detection.Secondly,the sea surface target detection method based on entropy feature is researched.First,the whitened spectrum is used in the doppler domain to achieve clutter suppression.Then,two entropy features,called relative sample entropy and discrete relative entropy,are proposed.The former can effectively distinguish the difference between the spectrum sequence of the sea clutter and target echo,and the latter can further realize the clutter suppression in the entropy domain,which has the characteristics of dual clutter suppression.Finally,the experimental results show that the proposed two detectors can significantly improve the performance of small target detection and ensure the robustness of the performance.Thirdly,the sea surface target detection method based on multi-domain and multidimensional features is researched.First,the complementarity of seven features in time domain,frequency domain and time-frequency domain is analyzed,and then the seven-dimensional feature space is constructed.Then,a binary classifier based on e Xtreme Gradient Boosting(XGBoost)with dual false alarm control is proposed to solve the problem of designing highdimensional classifiers.In the first part,the loss function of XGBoost is redefined to obtain the rough false alarm control at the structural level.In the second part,the classification probability is used as a statistic to obtain the precise false alarm control at the parameter level.Finally,experimental results on measured data show that the proposed detector can accurately control false alarm and has robust detection capability compared to the existing high-dimensional feature detectors.