Abstract: In order to solve the target imaging problem of the ultra-wideband(UWB) through-wall radar, a technique based on the combination of back projection algorithm and support vector machine is proposed. In this technique, data for imaging can be obtained by using BP algorithm and are classified by the support vector machine(SVM). It can be employed for positioning and recognition of the targets behind the wall. The simulation results based on the data from the through-wall detection model verify the feasibility and validity. Furthermore, the results also demonstrate that the spatial resolution is very high and the shapes of the targets which have different shapes can be reconstructed by using this approach. In addition, when the data are corrupted by the noises, the shape of the targets behind the wall can still be well predicted, which means the robustness of the technique. Finally, the classification accuracy analysis for different sampling lengths and sampling intervals show that the influences of the sampling lengths and sampling intervals on the shape recognition are limited, and that the classification accuracy can be improved with the increase of the sampling lengths and the decrease of the sampling intervals.