Abstract: The regularization term for semi-airborne transient electromagnetic regularization of long-line source usually adopts L2 norm, and the fitting result is relatively smooth, which cannot effectively describe the layer interface information. Aiming at the stratified medium steep change model to realize the inversion algorithm whose regular term is the L1 norm, the authors transform the original problem into the L2 regularization sub-problem by the iterative re-weighted least squares method to solve the problem of non-differentiation in the L1 norm; OpenMP technology is used to solve the problem. The parallel calculation of the Jacobian matrix improves the inversion speed; the adjustment strategy of the adaptive regularization factor segmentation iteration method is analyzed and improved. The improved adaptive regularization factor adjustment strategy is more suitable for semi-airborne transient electromagnetic inversion algorithm of L1-norm regularization. Finally, the resistivity is inverted and compared with the Occam inversion results. The results show that the inversion of L1-norm regularization can highlight the electrical interface conforming to the real model after sufficient iterations, and the inversion resistivity is closer to the true value of the model.