Turbulence intensity profile can be obtained by turbulence intensity profile lidar system through measuring a set of atmospheric coherence length. However, direct inversion from measured atmospheric coherence length data can result in a high noise gain. A new approach, which combines slope and Richardson method, was developed and its numerical simulation were performed. By recovering two representative Hufnagel-Valley(H-V) C2n profiles, the average relative errors between initial profile and recovery profile are 7.8% and 10.6%, respectively. Numerical results demonstrate that this method has a good precision.