Due to the complex tunnel construction environment, random value pulse noise is a common and difficult to eliminate interference in tunnel advance prediction. Random value pulse noise will cause serious interference to the detection signal, resulting in the failure to accurately obtain the lithology information of underground environment in tunnel construction area, so it is very necessary to suppress random value pulse noise. In this paper, morphological filtering algorithm is applied to suppress random value pulse noise interference in TSP data. Morphological filtering is a powerful tool to extract shock features. According to random value pulse noise characteristics in seismic records, the open and close operators in morphological filtering algorithm can effectively suppress positive and negative pulse noise in seismic records. The results of seismic data simulation processing and analysis show that the mathematical morphological filter can effectively suppress the interference of random value pulse noise, greatly improve the data quality of reflected P wave and reflect S wave. The method presented in this paper is of great significance not only for tunnel geological prediction, but also for improving the accuracy of seismic exploration and improve the distance and accuracy of tunnel geological advance detection.