In the field of civil engineering, the survey of the thin top layer of carriageways, i.e. the pavement layer, is hardly performed by the conventional radar Non Destructive Techniques. This thesis aims to improve the time resolution of the classical GPR radar by a factor of 3 at least, using super and high signal processing techniques applied to time delay estimation (TDE). At first, the most well-known super and high resolution techniques (LSMYW, MUSIC, ESPRIT, Min-Norm) are studied and evaluated on simulated data. The performances of the algorithms are assessed in terms of the root mean square error on the estimated thickness of pavement material. An accuracy criterion on the measured thickness allows to deduce the ultimate time resolution of each algorithm. For highly or totally correlated echoes, the sub-band smoothing techniques are optimized. Whatever the correlation magnitude, the performances of algorithms are sufficient to fulfil the required criterion for the NDT application under scope. Then, among the subspace algorithms, the study focusses on those which require a lower computational cost. The ESPRIT algorithm is adapted and extended to the signal model with the radar pulse and any noise characteristics accounted for. Moreover, two linear subspace algorithms (OPM and SWEDE) are introduced. A polynomial solution is proposed for OPM (root-OPM) which is more efficient to perform the TDE. SWEDE is reformulated to the TDE problem, and a more efficient whitening technique is proposed to improve its performances (FBD-SWEDE). At last, the principle of both SWEDE and ESPRIT algorithms are combined to develop three new algorithms applied to the TDE of radar signal (ESPRITWED, AV-ESPRITWED, G-ESPRITWED). Finally, the algorithms are tested on radar data, obtained from measurements on one layer and two-layers media. The implementation of the algorithms is detailed for both types of GPR devices (step-frequency and pulse radars). The results are found in accordance with the simulated results and allow us to conclude on the feasibility of the thin pavement thickness measurement from a 2 GHz radar bandwidth.
