摘要
本文针对时间推移地震本身包含不同时期的两次或者两次以上的勘探反问题,构造了一种快速有效的反演方法——多尺度全变分法.通过引入全变分正则化来代替传统的Tikhonov正则化,针对待反演参数不连续的情况,提高了算法精度.为了提高计算效率,引入了多尺度方法,从而构造了多尺度全变分方法.在数值模拟中,针对一个时间推移地震反演问题对多尺度-Tikhonov正则化法、单一尺度全变分法、以及本文所构造的多尺度全变分法进行了比较.结果表明,本文所提出的多尺度全变分法是一种稳定、快速和精确的反演方法.
In this paper, a multiscale total variation method is applied to parameter identification problems for time-lapse seismic which consists of two or more seismic surveys shooting at different calendars. For the discontinuous parameters, a total variation regularization method is introduced instead of the traditional Tikhonov regularization. Combined with a multiscale method, a multiscale total variation regularization method is put forward to improve computational efficiency and accuracy. In numerical simulation, the multiscale total variation regularization method is compared with both the multiscale-Tikhonov regularization method and a single scale total variation regularization method for time-lapse seismic inversion. The numerical results show that the proposed multiscale total variation method is more stable, faster and more accurate than other methods.
出处
《地球物理学报》
SCIE
EI
CAS
CSCD
北大核心
2010年第8期1883-1892,共10页
Chinese Journal of Geophysics
基金
国家自然科学基金项目(40774056)
中国博士后科学基金项目(20080430524)
哈尔滨工业大学优秀青年教师培养计划(HITQNJS2008032)资助
关键词
多尺度全变分法
全变分
时间推移地震
反演
Multiscale total variation method, Total variation, Time-lapse seismic, Inversion
