| Ground Penetrating Radar (GPR) technique is an advanced detection method of subsurface target. Exact information of subsurface target can be obtained only by system calibration because of the Ultra-wideband characteristics of GPR system and the complexity of electromagnetic waves propagation in breakthrough detection. This paper mainly studies System Calibration which is a key technology.Firstly, in both the radar system and environmental medium, various factors which have influence on the detection performance are analyzed, especially the transmitter, antenna, receiver and dispersive medium. All this influence factors can cause some problems including the waveform distortion, bandwidth narrow, waveform dither and etc. It is directly affects the precision, resolution and reliability of the system.Secondly, based on the analysis of various influence factors and GPR System Calibration's aim and assignment, a new approach is proposed, which calibrates the radar system, medium and target as a whole. Traditional System Calibration calibrates radar system and medium separately. Integral System Calibration differ from the traditional method, which can simplify the process, improve the precision of calibration. Meanwhile, the system mode of GPR is built and the principle of the Integral System Calibration.Thirdly, in order to eliminate the radar echo's distortion and improve the resolution of GPR, various common deconvolution methods are studied. Because the Blind-Deconvolution can estimate the radar wavelet and reflection sequences when the radar wavelet is unknown and not be restricted by the phase of radar wavelet, it is more suitable for the Integral System Calibration. But it has the disadvantages of poor applicability and large calculation.At last, by further studying the MCMC method, a blind-deconvolution system calibration method based on the MCMC is proposed to complete Integral System Calibration of GPR. MCMC is suitable for the non-linear non-Gaussian mode and simplify the calculation. Combined with the Blind-Deconvolution, the new approach is not easily influenced by the transmitting signal, it still has the fast calculation speed and can estimate the radar wavelet and reflection sequences accurately by the introduction of prior information. The simulation experiment results prove that this new algorithm is effective. Additionally the EM approach based on the Maximum Likelihood is studied. By comparing experiment results under different SNR, it is shown that system calibration method based on the MCMC has the faster convergence speed, better anti noise performance and higher calibration precision. |
PDF Full Text Request