Research On Intelligent Target Detection Method In Sea Clutter Environment

In modern national defense,sea detection by radar is the main way to realize sea surveillance and provide electronic information.It is of great value to study radar target detection in sea clutter environment.The research in this field faces two main challenges: first,the target signal-to-clutter ratio is not high;second,the performance of traditional detector is limited due to the mismatch of statistical model.In recent years,with the deepening of theoretical research and the improvement of hardware computing level,artificial intelligence technology represented by deep learning method has become a potential solution for sea exploration technology.Focusing on radar target detection in sea clutter environment,this paper mainly studies sea clutter suppression technology and target detector by using intelligent methods based on sea detection working scenes.The main contents are as follows:1.The radar signal model and the common statistical model of sea clutter are given according to the working scene of radar sea detection,and simulation experiments are carried out.There is a gap between the statistical characteristics of sea clutter and the actual sea conditions,so the measured sea clutter data are obtained and analyzed.On the basis of sea clutter data,the signal-level processing,such as adding cooperative target and pulse compression,is carried out to provide support for the subsequent research on data-driven methods.2.To solve the problem of low target signal-to-clutter ratio,a sea clutter suppression technology based on deep learning is proposed.Combined with the idea of segmentation,the structure of sea clutter suppression network is designed by using an improved U-NET framework.In order to adapt the proposed technique to the sea clutter suppression task,a reasonable label is designed and normalized on the data set.When the training is supervised,the optimization method of RADAM +SGD is proposed to make the proposed network converge faster and better in the training.In the working stage,the proposed method is initially verified to enhance the target SCR effectively.Two evaluation indexes were designed and repeated experiments.When SCR reached6 d B or above after pulse compression,the inhibition rate of the proposed network reached at least 0.988,and the target was enhanced at least 23.79 d B in the case of successful inhibition.The proposed method is suitable for multi-target and strong clutter scenarios and has certain robustness.3.A data-driven intelligent detector was proposed to solve the problem of limited performance of traditional detectors due to the mismatch of statistical models.Based on the end-to-end idea of deep learning,an intelligent detector structure composed of Densenet feature extraction module and FC prediction output module is designed.Through data set design and detector training,the intelligent detector has the ability of detection and decision.In the experimental stage,the limitations of CFAR detection in the measured sea clutter environment were verified from two perspectives: false alarm control and target detection.The performance of the intelligent detector was evaluated by Monte Carlo test.The comparative test shows that the detection probability of the intelligent detector is higher than that of the CFAR detector.There are two main innovations in this paper.First,a data-driven sea clutter suppression method is proposed to effectively suppress sea clutter and improve the sign-to-clutter ratio at the cooperative target.Second,a data-driven intelligent detector is proposed to achieve a higher detection probability than the traditional method in the sea clutter environment.