Researches On Lightning Signal Detection Methods And Related Issues

This dissertation focuses on the issue of CFAR lightning signal detection which is an important component in the lightning location system. Based on the characteristics analysis of lightning, two CFAR detection algorithms for lightning signal are proposed. Several methods of CFAR detection are discussed and proposed to improve the performance of the existing detectors. Computer simulations and their results are given to verify the efficiency of these proposed methods. The main contributions of this dissertation can be concluded as follows:1,Lightning data are gatherd synchronously by a lightning signal acquisition and real-time streaming storage system at two different VHF bands. Based on short-time energy method, the time-domain characteristics of electromagnetic radiation signal of lightning discharge is obtained by statistical analysis on the gathered data. According to the spectrum analysis at these VHF bands, it can be concluded that the lightning signal at 290MHz is smaller interfered by background noise than the other band. The gathered lightning data is analyzed by the statistical methods such as mean, variance and kurtosis. It is clear that variance and kurtosis can be used to detect the lightning signal and their statistical properties are both affected by the length of block data.2,Since the occurrence and duration of lightning both are random and the background noise changes all the time, by the combination of the auto-filter technique and Excision detection, a novel CFAR lightning signal detection algorithm based on the short-time energy is proposed. However it has a phenomenon that the signal may be submerged by the noise when the length of signal is smaller than that of block data. By the mathematics analysis on the definition of variance and kurtosis and by the simulation on the statistics of the gathered data, it can be concluded that the energy and kurtosis has a complementary merit used in detection of block data. In order to improve the detection probability a joint detection algorithm is proposed and simulated by gathered data. Simulation results demonstrate that the joint detection algorithm is better than the short-time energy detection methods. 3,As the routine technologies of CFAR detection are not robust under different background environments, based on the statistics of VI and MR, a modified Vl-CFAR is proposed to improve the detection performance, in which the OS-CFAR of IVI-CFAR detection is substituted by OSCA-CFAR and OSSO-CFAR. Based on the research of the VI-CFAR detection algorithm and its improved algorithms, it can be found that these algorithms have the drawback of large processing time for order statistics operation and that the performance of those detection algorithms is not optimal. So a modified VI-CFAR detection is proposed to settle these problems. Simulation results demonstrate that MVI-CFAR detection is better than other detection methods. The MVI-CFAR detection algorithm has the advantage of low order statistics operation, and it can be used to settle random appearance of the interfering targets and the supreme number of the interfering targets in the single reference window. Lastly it can be found that all kinds of VI-CFAR detection have the drawback of the supreme number of the interfering targets.4,A novel ATM-CFAR detection algorithm is proposed, which is the combination of the monotony for the variable ODV and TM-CFAR detection algorithm based on the block method. The ATM-CFAR detection algorithm can overcome the shortcoming of the supreme number of the interfering targets that MVI-CFAR detection and OS-CFAR detection have, and it can improve the controlling capability of flase alarm rate at the strong clutter edge and reduce the computational complexity for the ACCA-ODV detection, which is based on the combination of the variable of ODV and ACCA technique. By simulations and comparing with other detection methods, the validity of ATM-CFAR detection is testified. The main advantages of ATM-CFAR detection can be concluded as follows:Fisrtly, the detection performance of ATM-CFAR is close to the CA-CFAR detection under homogeneous background. Secondly, its supreme number of interfering targets is increased comparing to the MVI-CFAR detection and the OS-CFAR detection. Thirdly, its controlling capability of flase alarm rate is better than the ACCA-ODV under stronger clutter edge environment, and more robust in practice. When using parallel block method and two-level architecture to implement detection algorithm, the ATM-CFAR detection needs less harware resource than that of ACCA-ODV detection. And it also yields a considerable save in computational complexity.5,Two kinds of modeling methods and the estimation of covariance matrix are studied under the non-homogeneous environment of the Gaussian distributed clutter, which is based on the knowledge-aided theory and uses the complex Wishart distribution and the inverse complex Wishart distribution. Then a modeling method about non-homogenous environment of Compound-Gaussian on clutter distribution is proposed, which is more consistent with the operational environment in practice. Under this model, the MMSE estimation of covariance matrix is given by the method of Gibbs sampling and Bayesian theory, and then a novel ANMF detector is obtained by substituting this estimation of covariance matrix in ANMF. The convergence of Gibbs sampling is studied and a few related parameters are obtained by computer simulations. Then the performance of ANMF detector is analyzed by using these parameters, and it can be known that the proposed detector has CFAR characteristics. Lastly by comparing to the other detectors, the novel ANMF detector is valid and optimal.