Abstract: The ground penetrating radar (GPR) can be used to detect and determine the scale, shape and position of hidden karst caves in tunnel construction, and it is very important for the protection of the tunnel construction safety and the hazard geology treatment. Due to the complexity of tunnel detection environment, the location calibration and the shape determination of the results for the traditional GPR 2D detection are difficult. However, due to the strong subjectivity of amplitude threshold setting, there is great uncertainty in the visualization process of GPR 3D data obtained based on the multiple survey lines. A 3D visualization method for the tunnel karst cave of GPR data is proposed. Firstly, to improve the imaging accuracy of karst cave targets, the F-K method is used to process each GPR B-scan. According to the coordinate information of GPR data, the GPR 3D data of the karst cave is synthesized to enhance the horizontal connection between different lines. Then, to improve the view effects and enhance the contrast of the effective reflection data, the method of GPR attribute analysis is used. The amplitude threshold of GPR 3D visualization of hidden karst cave is further extracted by using the K-means cluster method. Finally, the GPR 3D visualization of the tunnel karst cave can be realized by combining the attribute volume and the isosurface extraction technology. The reliability and adaptability of this method are verified by the model tests and field case analysis. The maximum spectral amplitude attribute is the optimal attribute of GPR signal in the proposed method, which may improve the radar view effect and enhance the contrast between the background and the effective reflection for GPR data. Furthermore, the proposed method solves the problem that the amplitude threshold setting of GPR 3D visualization excessively depends on the experience judgment of interpreters, and the results will be valuable for the stratigraphic analysis such as sedimentary strata.