Study And Application Of Wavelet Transform In Filter And Migration Of GPR

Ground penetrating radar is a geophysical exploration method which detects the underground strcture and characteristics. Because of the advantages such as high efficiency, quick speed, undamaged, and strong anti-noise features, it has been applied to every aspect in engineering domain, and it is a powerful tool of shallow layer exploration. Precision of GPR exploration has important connect with data processing, processing technique which is not fit is negative to detecting result directly. With information technique and electron technique rapid development, collecting data is more various, traditional GPR processing technique is not obviously sufficient. This thesis studied the filter and migration techniques for GPR.To GPR filter technique, according to noise's difficient features thesis has respective studied. Because of the features of direct arrive wave that has strong correlation, strong power, horizontal, first arriveal, to this tranditional filter method is KL transform, it is finded that KL transform is only description to information of record trace's whole time axis after study KL transform's theory, not to time axis's the information, namely while KL transform filter has improved image's horizontal resolution, there is interference in vertical resolution to reduce of vertical resolution. Combining with features of wavelet's parted time and frequency, it is better to resolve part questions. Therefore, this thesis bring forward wavelet's KL transform to filter direct arrive and horizontal wave, this algorithm is not only more improve horizontal resolution than KL transform, but also overcome disadvantage of KL transform in vertical axis that is not interference.Because of the features of random noise that has random, not regulation, tranditional filter method is introduced from seismic processsig techniques, but GPR signal belongs to instantaneously unsmoothed signal. Contradiction that was along wavelet's features limited it's utility after analysed wavelet's theory, multiwavelet can resolve this question, whereas multiwavelet need prefilter before its transform, prefilter might result to imbalance along input multitrace which is limit multiwavelet's superiority, so balance is necessary. Whether chosed threshold is fit that is vital to filter results. This thesis bring forward algorithm of improved threshold combining with traditonal algorithms. Experiments show that there is good filter effect to use new algorithm. This thesis put forward balance multiwavelet and algorithm of improved threshold to filter random noise. Models and practical applications show that this algorithm have better effect than traditonal ones.To GPR migration technique, the author firstly analyzes the tradition methods'question that Kirchhoff's has large noise and obvious arc, finite difference method has obliqueity limitation and lowly calculation efficiency, F-K migration has advantages of finite difference and Kirchhoff methods, but its disadvantages is sensitive to various velocity. Thereby, this thesis put forward multi-wavelet's F-K migration combining with which multi-wavelet's horiaontal coefficients of high frequency can hold true signal after decomposion. Experiments show that this algorithm can better extract diffraction wave for that it is important to actural migration diffraction wave. Finally, this thesis use improved algorithm of filter and migration on models and practical records to processing and analysis.