【摘要】 传统的匹配追踪方法通常采用完备的时频原子库和贪婪算法进行子波分解,计算量巨大。为此本文引入互相关阈值约束技术,实现两次迭代和完备库之间的反馈,剔除那些互相关值很小的时频原子,动态地减小完备库的冗余度;通过多原子同步提取,实现快速匹配追踪地震道分解。分解后的地震信号是按子波的能量排序的,每个子波都由振幅、频率和中心时间三参数表示,代表原始地震信号在某一中心时间的局部主振幅特征和主频率特征,因此在信号重构时,可根据有效信号与噪声信号的主振幅、主频率和中心时间的差异,构建三参数时频特征滤波器,筛选出代表有效信号的子波对信号进行重构,达到无损去噪的目的。合成地震记录和实际资料处理结果表明了该方法的有效性。更多还原

【Abstract】 Matching pursuit (MP) algorithm can adaptively decomposes a seismic trace into a series of constituent wavelets. Conventional matching pursuit decompositions usually need a sufficiently large atom dictionary and greedy algorithms to decompose wavelet with timeconsumed calculations. Therefor,we introduce a thresholdconstrained crosscorrelation approach. This approach implement a feedback loop between the iterations and the dictionary,and atoms with correlation ratios below the threshold are rejected. It can selfadaptive reduce the size of dictionary. Fast matching pursuit seismic trace decomposition is achieved through multiple atom extractions. After decomposition,each wavelet sorted in energy was expressed by three parameters,amplitude,frequency and centre time,which represents major amplitude and dominant frequency of original signals at a centre time. According to the difference of the major amplitude,dominant frequency and centre time between signal and noise,we construct a three parameter filter,and extract the signal wavelets to reconstruct the seismic trace to filter noise without signal damage. Tests on synthetic and real data show the effectiveness of the proposed method.更多还原