The first fundamental step within a research project aiming at the 3D reconstruction and rendering of the accident scenes involving motor vehicles is accounted for. More specifically, the paper deals with the problem of accurately locating and georeferencing both the considered scenario and all involved elements (such as cars, debris, peeling out and so on) of interest for Road Safety experts. A new method, based on Kalman filtering, is presented for enhancing the preliminary estimates, obtained through low-cost GPS devices, of the position of proper scene markers. To this aim, the distances between the markers is measured through a low-accuracy device (as an example, an ultrawideband radar) and the achieved results are used as feedback to improve the estimate of markers positions. Preliminary results, obtained in numerical tests, are presented to prove the feasibility of the method, thus suggesting its implementation in an actual measurement station.