To ensure the transverse continuity of the inversion results of central loop transient electromagnetic (TEM) data, it is practical to combine multiple survey points into a whole by applying transverse constraints and to establish correlations between adjacent survey points. This study employed the constraints of weighted transverse and vertical roughness to improve the correlations between adjacent survey points and thus to improve the inversion accuracy. Furthermore, this study introduced the adaptive regularization factor, which was obtained adaptively by calculating the norms of the data and the model objective functions in each iteration. In this manner, the dependence on the initial model can be reduced. Finally, using the algorithm developed, this study conducted the inversion of the measured data and data from three-dimensional forward modeling based on the unstructured time-domain finite element method. The inversion results are consistent with the actual results, indicating that the proposed algorithm in this study can enhance the continuity of the inversion results and has a high anti-noise ability.