SEPARATION OF BLENDED DATA BY SPARSE INVERSION BASED ON THE RECIPROCITY THEOREM

摘要

Recently a great cnange or mindset in seismic acquisition has occurred and much attention has been drawn to blended acquisition. Different sources at several locations are shot in an overlapping fashion and blended records are acquired, so that acquisition efficiency and potentially image quality can be significantly improved. The key factor of Wended acquisition is source separation, which is the procedure of recovering data as if they were acquired in the conventional survey. From the mathematics point of view, it can be formulated as solving an underdetermined equation such as Ax = b, yet a simple least-square inversion is only able to get pseudodeblend data where the blending noises are not able to be removed. In this paper we find that in the 2D regularization acquisition system, the blended data in common receiver domain can be simply connected with that in common source domain according to the reciprocity theorem, so that the above equation would be more determined to be solved. While in the implementation procedure, spgL1 norm basic pursuit sparse inversion algorithm is utilized calculating the reflection coefficients of the single source data, and then the deblended data are acquired via convolution. Ideal results could be produced in field data tests.