Full Waveform Inversion (FWI) is a very general multi‐parameters quantitative imaging method originally developed to obtain high resolution images of velocities and attenuation in the natural underground media. FWI promises interesting performances for civil engineering applications like cavity detection. Performances of the FWI method in seismic exploration are difficult to quantify because in real experimentations, the properties of the media are not known. Furthermore, characteristics of the source and coupling of receivers are not controlled. In order to quantify the performances of FWI and its adaptability to near subsurface applications, small scale physicals models are realized and a non‐contact laser ultrasonic laboratory is dedicated to simulate seismic reflection measurement configurations at ultrasonic scale. Seismograms well reproduce real scale data in terms of waveforms and repartition of energy. A FWI algorithm developed by Geosciences Azur and applied in geological investigation is used. The direct part of the algorithm uses a frequency domain finite difference modelisation. The inverted model is obtained by using a Gauss‐Newton method of optimization. Results of inversion of multi‐receivers and multi‐sources data obtained from simple physical models are presented.