Research On Moving Target Indication Method Based On Pattern Recognition

Ground moving target indication can be used for target detection and motion parameter estimation.It is widely used in military and civilian field.However,when airborne radar detects ground the moving target,there are serious background interference,including clutter,jamming and noise.How to accurately estimate the moving target parameters from the interference has always been the direction of researchers.With the recent development of high speed,high performance digital signal processors,space time adaptive processing has been applied to ground moving target indication and has become an important part of airborne radar for moving target indication functions.However,STAP has some shortcomings.In order to solve these problems,with the rise and development of emerging technologies,researchers have a new direction.In recent years,with the improvement of pattern recognition,pattern recognition has become more and more mature in the radar field.Recently,a learning method based on moving target indication was proposed by researchers.The restrictions of STAP methods on the enough range gates of secondary data are no longer necessary,but there are still many difficult problems in moving target indication.This thesis focuses on the random phase target echo,the small amount of secondary data,the heterogenous environment and the low signal clutter ratio scenario involved in moving target indication,and the thesis carried out theoretical research,synthetic simulation and experimental verification.The main innovations and contributions are as follows:1.A simplified model is researched for signals received by airborne pulsed Doppler phased array radars,including moving target model,noise model,jamming model and clutter model.Then the estimation method of covariance matrix of interference plus noise is presented.The traditional fully adaptive STAP algorithm is analyzed.The process of STAP based on classification is given.2.Due to the fact that the required secondary data of the STAP based on polynomial classification method is still more to achieve good performance.The moving target indication based on support vector machine method is proposed in the thesis.Compared with the STAP based on polynomial classification method,the method can achieve good performance in the case of constant phase with only one secondary data.3.Because the STAP based on polynomial classification method and the moving target indication based on support vector machine method has poor performance in the random phase target echo model,the heterogenous clutter environment and the low signal clutter ratio scenario,the CNN-MTI method is proposed.Compared with the fully adaptive STAP method,the STAP based on polynomial classification method,and the moving target indication based on support vector machine method,the moving target indication based on convolutional neural network method can achieve good performance not only with very small amount of secondary data but also in the random phase target echo model,the heterogenous clutter environment and the low signal clutter ratio scenario.4.In order to address the problem of insufficient training data samples for moving target detection methods based on convolutional neural network method,the thesis proposes to augment the training dataset by taking the secondary data as the background interference and combining them with the artificially generated moving target signal at different signal clutter ratio levels,different Doppler frequencies and random phases.The above work has been verified by simulation data and Mountain Top experimental data.The results show that the above method can realize the aim of moving target indication under different conditions.