When there are water rich-karst caves or large faults in front of the tunnel face, blind construction may easily cause safety accidents such as water inrush, mud outburst and collapse. Using advanced geological prediction methods to predict the spatial distribution of unfavorable geological bodies can greatly reduce construction risks. This paper introduces a new seismic reflection imaging system TSP-SK, and discusses the features, working mode and data processing methods of the prediction system, and verifies the reliability of TSP-SK through practical tunnel engineering prediction case. The result shows that TSP-SK can collect high signal-to-noise ratio seismic reflection data with little interference from clean first arrival when the water sealing effect of explosives in the borehole is good. Good reflection wave pickup effect can be obtained when the maximum gain of inverse Q filter and Q value are taken in the range of 15 to 30. The prediction results of TSP-SK are in good agreement with the actual excavation results, that is, the areas with low P-wave velocity indicate poor rock mass quality, and the area of low S-wave velocity is more likely to produce water from construction.