Abstract: Strong rebar echo and its multiples will affect the performance of GPR（Ground penetrating radar）in disease target detection and discrimination badly, so they must be eliminated as much as possible. A method for rebar echo detection and suppression is proposed in this paper. First of all, GPR observed data are projected into Hyp-curvelet space by the Hyp-curvelet transform, and echoes from different targets are concentrated but separated with each other. Target can be detected by searching the peak values in Hyp-curvelet space. Rebar can be discriminated from disease target with initial phase of echo and features distribution in time-frequency domain, and then its echo can be eliminated in Hyp-curvelet space. In the end, data are reconstructed back into time-space domain without containing rebar echo. Simulation results show that the proposed method has good performance in rebar echo detection and suppression under the condition of low SNR, and diseases echoes are preserved well and rebar echo is rarely left over in the reconstructed data.