Magnetic flux leakage detection has received extensive attention and application as one of the main methods of non-destructive testing. The problem of identification of defects by magnetic flux leakage detection belongs to the inverse problem of electromagnetic field calculation. It is difficult to directly solve the problem due to its ill-posedness. Therefore, at present, most of the research and application of magnetic flux leakage detection methods are to avoid the search matching method which directly solves the inverse problem. The method of mixed regularization LSQR-Tikhonov is used to deal with the magnetic flux leakage detection inversion problem based on the magnetic dipole unit integration model. Add Tikhonov regularization to the LSQR method to obtain more main singular value information to reduce the error. The approximate solution obtained by regularization is closer to the true value, and the reflected two-dimensional inversion results can also determine the position and shape of the defect. The accuracy of the algorithm is verified by simulation and experiment.