The input data of reflection wave velocity tomography is the subsurface reflection information, which usually includes the reflection point of seismic section and the residual movement of common imaging gathers(CIGs). In the process, the key step is to pick intensively and automatically. To be specific, firstly structure tensor method is introduced to filter seismic data along structural features, so as to enhance the feature of the seismic section and the CIGs, improving the recognition of the reflection information. After that, the reflection information can be extracted more easily. In addition, based on the wavelet geometric feature of the seismic data and the spatial continuity of the seismic events, wave peak (trough) discrimination criteria can be established to pick the reflection points automatically, based on which the accurate reflection position picking results can be obtained. Finally, according to the reflection point pickup of seismic section, linear inversion method is introduced to calculate vertical shifts on each offset of the CIGs, which is used as the residual movement of the CIGs. The accuracy and effectiveness of this technique are verified by synthetics and field data.