Inverse of magnetic dipole field using a reversible jump Markov chain Monte Carlo (1310.0915v1)

Abstract: We consider a three-dimensional magnetic field produced by an arbitrary collection of dipoles. Assuming the magnetic vector or its gradient tensor field is measured above the earth surface, the inverse problem is to use the measurement data to find the location, strength, orientation and distribution of the dipoles underneath the surface. We propose a reversible jump Markov chain Monte Carlo (RJ-MCMC) algorithm for both the magnetic vector and its gradient tensor to deal with this trans-dimensional inverse problem where the number of unknowns is one of the unknowns. A special birth-death move strategy is designed to obtain a reasonable rate of acceptance for the RJ-MCMC sampling. Some preliminary results show the strength and challenges of the algorithm in inverting the magnetic measurement data through dipoles. Starting with an arbitrary single dipole, the algorithm automatically produces a cloud of dipoles to reproduce the observed magnetic field, and the true dipole distribution for a bulky object is better predicted than for a thin object. Multi-objects located at different depths remain a very challenging inverse problem.