Source localization techniques assume that elastic wave velocity distributions are homogeneous in many cases. However, the assumption would be violated if objects are locally and heavily damaged, and accuracies of identified source locations by the source localization techniques are consequently degraded on the heterogeneity. The authors proposed a source localization technique for overcoming the difficulty by adopting ray-trace technique to properly consider the heterogeneity of the elastic wave velocity distributions, and the relay point is introduced to raise accuracy of the identified source locations as well in the source localization technique. The source localization technique has been implemented in an algorithm of AE-Tomography and successfully contributed to identify elastic wave velocity distributions by using arrival times of elastic waves at receivers. The resolution of the identified source locations depends on densities of the relay points in the source localization technique, and it immediately implies that the relay points have to be finely installed in the objects of interest to raise the resolution of the source locations on objects of interest. Since the computational costs of the source localization technique depends on the total number of nodal and relay points on the objects, the computational costs drastically rise if it is intended that the source locations are identified in high resolution. The resolution of the identified source locations influences accuracies of identified elastic wave velocity distributions, and the degradation of the accuracies consequently degrades the accuracies of identified elastic wave velocity distributions. For overcoming this difficulty, the authors proposed a source localization technique that installs source candidates finely only in the vicinity of roughly estimated source locations. Nevertheless, this technique reduces the computational cost in comparison with the conventional source localization technique based on ray-trace techniques, it still expensive if the range of “the vicinity” is enlarged and the number of events is large. Thus, a new source localization technique is introduced in this study. The technique installs the source candidates finely in smaller region in comparison with the previously proposed technique. The region is moved and its size is changed adaptively in the computational procedure to raise the accuracy of the identified source locations. This procedure requires iterative computations for convergence. However, its total computational cost is reduced because the number of relay points is smaller and the cost for the ray-trace is reduced. The proposed technique verified by performing a series of numerical investigations and its capability is discussed. The results of the investigations suggested that the proposed source localization technique identifies the source locations of elastic waves accurately on the heterogeneous elastic wave velocity distribution. It is also confirmed that the accuracy of the identified source locations is controllable by changing parameters for the convergence and computational cost is reduced in comparison with the previously proposed technique.