Study On Radio Stealth Radar Signal Design And Recognition Method

The research of radar automatic target recognition(RATR) technology not only is of great theoretical significance, but also has a broad application prospect. Especially in military applications, the United States, Russia, Germany and other Western military powers also have put it as the key research content and key technology urgently required to be broken through for the future intelligent weapon system. Compared to the SAR/ISAR image, the HRRP is easy to be obtained and be processed, and the real-time target recognition is easy to be achieved for the data scale is not large. RATR based on HRRP has become a research focus at home and abroad. Meanwhile, with the rapid development of military science and technology, the modern radar encounters four kinds of threats, namely the "electronic interference", "stealth target", "anti-radiation missiles", and "low altitude/very low altitude penetration". Therefore, the radar radio frequency stealth(RF stealth) performance research has a very important strategic and application significance. This thesis focuses on the radio frequency stealth waveform design based on the HRRP recognition, and is mainly composed of two parts. In the first part, aiming at RATR application, two RF stealth RATR signal design methods based on the optimum transmit-receiver and the sequential hypothesis testing and based on the mutual information and the sequential hypothesis testing are proposed, which makes a beneficial exploration for achieving the radar RF stealth performance. In the second part, the angular-sector segmentation way and the kernel way are adopted for solving the HRRP target aspect sensitivity problem, and three recognition methods are put forward including grey system, the adaptive angular-sector segmentation and the maximal the margin kernel optimization. The main contents and contribution are as follows:1. The RF stealth RATR signal design based on the optimum transmit-receiver and sequential hypothesis testing is studied. The optimum transmit-receiver constitutes a closed-loop system of the transmitter, the external environment and the receiver, so it is a kind of the cognitive radar, which can adaptively adjust the radar signal and the corresponding receiver according to prior information. Meanwhile, sequential probability ratio test of sequential analysis theory not only makes two kinds of errors to be small enough, but also makes the sampling rate to be minimum. The sequential analysis theory is combined with optimum transmit-receiver, and the RF stealth RATR signal design method is presented. At first, this method ensures the recognition performance, and the RF stealth performance from two aspects of the radiation energy control and the RF stealth radar signal.2. The RF stealth RATR signal design based on the mutual information and the sequential hy-pothesis testing is studied. For the general radar system, the target impulse response does not always to be determined, and often is assumed to obey a random distribution. The Gaussian distribution is the most commonly used random. In this circumstance, the mutual information between the echo signal of the receiver and the target echo signal affects the radar system performance. The radar signal design method based on the mutual information is another kind of cognitive radar, which can transmit the suitable signal ensuring radar system performance. And the RF stealth RATR signal design method combined the mutual information design method and the sequential probability ratio test is proposed. This method can achieve the RF stealth performance on the basic of ensuring the recognition performance.3. The RATR based on grey system is studied. The HRRP is the vector sum of these corresponding scatters in the range resolution along the radar line of sight, and it reflects target structure information. The RATR based on HRRP is divided into the training stage and the recognition stage. In the training stage, the HRRP templates are formed through the comparison of similarity degree among these training HRRPs, and in the recognition stage the category of the test HRRP is determined by the similarity degree between the test HRRP and these HRRP templates, namely the maximum similarity degree corresponds to the target. The grey rational analysis is an important branch of the grey system theory, and its basic principle is that the relationship degree of system factors is analyzed through the comparison of the geometric relationship of system factors sequences’ curve. The more similar these curves are, the greater the relationships are. The target HRRP recognition method based on the grey rational analysis is designed. The method has lots of properties, such as clear physical meaning, its simple calculation, easy operation, etc.4. The RATR based on the adaptive angular-sector segmentation of the grey rational grade is studied. For the HRRP target aspect sensitivity problem, the fixed angular-sector segmentation way is simple and convenient, but it leads to the mismatch between the HRRP and the target scatter scattering characteristics, and then the recognition performance is decreased. The adaptive angular-sector segmentation adopts some angular-sector segmentation criteria according to the HRRP distribution or the classifier and adaptive achieve the angular-sector segmentation, so the method can solve the mismatch problem and increase the recognition performance. On the basic of the target HRRP recognition method based on the grey rational analysis, two RATR methods based on the adaptive angular-sector segmentation of the grey rational grade are presented, which adopt the grey rational grade to carry out the adaptive angular-sector segmentation, so these two methods can further improve the target recognition performance.5. The RATR based on the maximal margin kernel optimization is studied. Correspond to the angular-sector segmentation, the SVM kernel method is another method of solving the HRRP target aspect sensitivity problem. The recognition performance of SVM largely depends on the choice of the kernel function. The kernel function which is generally applicable to any data does not exist. The data dependent kernel function is a revised kernel function and can improve the generalization ability of the SVM classifier, but the objective function of the kernel optimization and the objective function of the SVM classifier are different, they successively correspond to the Fisher criterion and the maximum margin criterion. In order to unify the kernel optimization criterion and the SVM classifier solution criterion, the RATR method based on the maximal margin kernel optimization is designed. This method adopts the same optimization criterion, which is equivalent to optimizing the data dependent kernel function based on the structural risk minimization criterion for the kernel optimization based on the maximal margin principle and the SVM classifier solution adopt the same optimization criterion, so this method can improve the target recognition performance of the SVM classifier.