【作者】 王喜珍；

【导师】 高孟潭； 滕云田；

【作者基本信息】 中国地震局地球物理研究所 ， 固体地球物理学， 2004， 博士

【摘要】 以提升方法和劳伦多项式分解为基础，研究了基于整数小波变换的地震数据无损压缩方法。对传统的编码方法如HUFFMAN编码、ARITH编码、RUN-LENGTH编码和LZW编码来讲，除RUN-LENGTH编码外，小波变换都可以较大幅度地提高压缩比，其中LZW编码方法的压缩比最高，小波变换的效果也最好。与地震系统数据标准交换格式采用的Steim-1、steim-2编码方法相比，基于小波变换编码的压缩比较Steim-1编码的压缩比有显著的提高，较steim-2编码的压缩比也有提高。对噪声数据和实际观测地震数据的压缩结果表明基于小波变换的压缩方法在压缩比方面要优于其它的编码方法。小波函数对压缩比有一定的影响，对噪声数据的压缩结果表明CDF(2，)双正交小波组的压缩比最好，而对地震观测数据的压缩结果表明CRF(13，7)和SWE(13，7)两个双正交小波的压缩比最好。 利用双正交小波函数的优点，研究了地震数据有损压缩方法和重构误差对确定地震基本参数的影响。小波变换滤波器组系数长度和分解水平大小对信噪比的影响程度表明，当压缩比大于10，小波函数为bior(3，7)(B-样条双正交小波，其中3为重构滤波器系数长度，7为分解滤波器系数长度)、分解水平为5时信噪比较高，均方差较小。示例地震数据的压缩结果表明，压缩比为0～40时，信号重构误差对确定地震事件基本参数震级和震中矩的影响很小，几乎可以忽略。对重构信号的频谱和相位的分析表明，随着压缩比的增加，频谱和相位的变化增大，特别是压缩比约大于26时，二者变化相当明显。多个地震事件数据的压缩结果表明，当压缩比分别为10.5、21.0和30.5时，重构信号的幅值变化、拾取的P波和S波到时变化都较小，对确定震级和震中矩的误差在合理的范围之内。示例地震数据的压缩结果还表明，信噪比和均方差的变化存在一个相对平缓的区域，压缩比在该区域内取值，可以得到较好的压缩结果。 研究了震相到时拾取方法，提出了基于小波变换的特征函数构造方法。根据P波和S波震相的频率特征，小尺度上的小波分解系数对应信号的噪声部分，计算特征函数时可以不考虑这些系数。地方震、近震和远震的P波到时拾取结果表明，本文构造的特征函数较其它传统的特征函数拾取的到时精度有所提高。结果还表明在P波到时附近新的特征函数的变化梯度增大，对应的横轴范围变窄，更有利于P波到时的拾取。用实际记录的地震事件波形数据拾取到时的结果表明，STA／LTA和线性偏振两种到时拾取方法在经小波分解和没有小波分解时确定的P波到时有较大的差别，基于小波分解的到时拾取方法确定的P波到时精度更高，对确定震中矩的影响也最小。更多还原

【Abstract】 The lossless compression of seismic data was studied on the basis of the integer wavelet transform, which was based on the theoretical analysis of the Lifting scheme and the Laurent polyphase matrix representation. To the traditional code method like the HUFFMAN code, the ARITH code, the RUN-LENGTH code and the LZW code, except the RUN-LENGTH code, it was found that the compression ratio was largely increased by wavelet transform. And it was also found the compression ration of the LZW code is the highest and the effect of the wavelet transform is the best. Comparing to the Steim-1 code and Steim-2 code of the standard exchange format of the seismic data, the compression ratio based on the wavelet transform is much higher than on steim-1 and little higher than on steim-2. The compression results of the noise data and the actual observation seismic data show that the compression code on the basis of the wavelet transform is better than the other kinds of compression code. The wavelet function affects the compression ratio to some extent, the compression results of noise data told us the compression ratio of the biorthogonal wavelet function CDF, is best, while to the result of the observation seismic data, it shows the compression ratios of the CRF(13,7) and SWE(13,7) biorthogonal wavelet function are better.Based on the biorthogonal wavelet function merit, it is studied the lossy compression of seismic data and the influence of the reconstruction error on the calculation of the basic earthquake parameters. According to the influence of the coefficient length of wavelet transform filter group and the size of decomposition level on the S/N(the ratio of signal and noise) and RMSE(the root of mean square error), it is believed that the S/N is higher and RMSE is smaller at the condition that the compression ratio is lager than 10 and the decomposition level of bior (3,7) is 5, where bior(3,7) is B-spline biorthogonal wavelet, and 3 is the coefficient length of the reconstruction filter and 7 is the coefficient length of the decomposition filter. From the compression results of the seismic data as the given example, it is known when the compression ratio is in the range 0~40, the signal reconstruction error has little affect on calculating the magnitude and the epicenter distance of the earthquake event, even can be neglected. The analysis of the reconstruction signal frequency spectrum and the phase indicates the change of the frequency spectrum and phase increase with the increase of the compression ratio. Specially, when the compression ratio is approximately larger than 26, their changes are much obvious. The compression results of many earthquake event data show the changes, which include the magnitude, picking onset time of the P wave and S wave of the reconstruction signals, are smaller at the compress ratio 10.5, 21 and 30.5 respectively, and the error of the seismic magnitude and epicenter distance is at the reasonable scope. The compression results of the given example also told us that the change of the S/N and the RMSE has a relatively flat region, and it was thought the compression ratio is quite ideal in this region.On the basis of wavelet transform, it was studied the method of picking onset time of seismic phase and proposed a new method for constructing the characteristic function. According to the frequency features of the P wave and the S wave, we think that the wavelet decomposition coefficient at the smaller scale is corresponding to the noise and should be omitted when calculating the characteristic function. Studying on the local earthquake and remote earthquake, the results show the precision of picking onset time of P phase is higher for the new characteristic function than the other traditional ones. The result also indicated near to theonset of the P wave, the gradient of the new characteristic function increases and the corresponding abscissa axis scope is narrower, so it is benefit to picking P phase onset time. From result tested by the real earthquake event data, it was found the更多还原