In order to improve the recognition accuracy of the underwater riprap, the physical model and resistivity tomography model of the riprap with internal cavities are established. The performance of resistivity tomography technology in the detection of internal cavities inside the underwater riprap is studied and compared with the traditional Kriging interpolation. The results show that the resistivity tomography can effectively identify the internal cavity in the underwater riprap and the inversion accuracy reaches a higher value when the number of stimulation points increases to 6. The identification ability of the resistivity tomography on the longitudinal section of channel is much higher than that of the traditional Kriging method. Meanwhile, if the region with conductivity greater than 300.μS/cm in inversion results is regarded as a cavity region, the overall inversion results are slightly larger than the real situation, and the deviations in the three directions of x、y、z are 2.73, 1.75, and 6.95.cm, respectively. Based on the inversion results, the forward calculation of the resistivity tomography model is carried out. It can be found that the fitting degree between the forward modeling and the measured data was 0.95, which proved the accuracy of the resistivity tomography inversion.