Abstract: The traditional singular value decomposition (SVD) method decomposes the transfer coefficient matrix in one go, which fails to fully consider data changes and is sensitive to noise interference, resulting in limited reconstruction accuracy. This article proposes a radiation source reconstruction method that combines Adaptive Singular Value Decomposition (ASVD) and Tikhonov regularization. This method dynamically adjusts the SVD decomposition by iteratively decomposing the transfer coefficient matrix and setting a threshold based on the maximum singular value, effectively suppressing numerical errors caused by small singular values. At the same time, this article systematically analyzes the effects of regularization parameter selection, \boldsymbolP_\boldsymbolz dipole contribution, Gaussian white noise interference, and frequency variation on reconstruction accuracy. The simulation and actual measurement results show that the relative error between the reconstructed amplitude component of the proposed method and the reference result is controlled below 2.6%, which verifies the good accuracy of the method in radiation source reconstruction.