Traditional algorithms utilized for traveltime tomography are not well suitable for handling large 3D seismic data because of the shortcomings inherited in the ray tracing technique. Besides, the slow convergence speed of classical algorithms based on Tikhonov's regularization are also unacceptable as only few iterations are affordable in real large 3D problems. In this paper, we develop an efficient 3D first arrival travel time tomography scheme based on the upwind finite differences algorithms and the shaping regularization. No cumbersome ray tracing techniques are needed since the forward and adjoint problems can be solved efficiently by using several linear operators we proposed based on the upwind differences technique. Moreover, the speed of convergence can be improved considerably with the usage of shaping regularization. The numerical experiment on 3D Marmousi model has demonstrated the better accuracy and faster convergence of proposed method compared with classical algorithm.
展开▼
