Research On Signal Processing Algorithm Of Ground Penetrating Radar

Ground Penetrating Radar (GPR) is a no-wasting type of explorational technique, which is developing rapidly in recent years, and it belongs to radar technique's application field which is new and developing increasingly. GPR can be widely used for media (for example rock and soil) exploration, configuration of project and construction exploration, entironment detecting, against terror and so on. This dissertation focuses on the two main applications of GPR: Linear Frequency-Modulate Continuous Wave (LFMCW) GPR and Impulse (IP) GPR. The focuses in this thesis are to suggest and evaluate algorithm of underground medium parameters inversion for LFMCW GPR and layer picking method of underground medium for IP GPR. The main contribution of this research can be summarized as follows:(1) Analyzing plane wave propagation phenomena and echo propagation characteristic. Basing on forward modeling theory of convolution method, we set up a forward modeling of LFMCW GPR and then typical GPR echo data is simulated. It establishes the base for the research of inversion algorithms.(2) In the research of underground medium parameters inversion algorithm of LFMCW GPR, it includes traditional electromagnetism inversion algorithms: layer stripping inversion method and Gauss Newton inversion method, and their localization is analyzed. In order to inverse underground medium parameters more effectively, an inversion algorithm for underground medium parameter estimation based on MUSIC spectrum estimation is put forward in this dissertation. In this algorithm, after estimating the time delay and amplitude of LFMCW GPR echo, the accurate medium parameter is worked out and then the right medium structure underground is gotten in succession. The simulating results show that the new algorithm improves much on the precision comparing with traditional electromagnetism inversion algorithms.(3) In the research of underground medium layer picking method of IP GPR. At first we research the layer picking based on hidden Markov model and analyze its localization. In order to repair its shortage when the method carries out layer picking, a layer picking based on hidden Markov model and Bresenham algorithm is presented in this dissertation. In this method, after tracking time of delay (TOD) of GPR echoes, edge detection of layer is accomplished primarily. Basing on it, edge linking is come true and layer picking is accomplished at last. The results of dealing with actual data show that the layer picking method in this paper improves much on the precision of tracking comparing with layer picking only based on hidden Markov model.