Abstract: To detect low resistivity abnormal bodies in rock strata and prevent mine water hazards, four relevant geoelectric models are established by using Visual Fortran software. Dipole devices are selected to imitate the pole-running mode of actual field data acquisition. Forward modeling based on finite element theory is carried out for the self-modeling model, especially with the aid of water layer inversion. Stereo imaging and multi-directional slicing are used to process three-dimensional data volume. The differences and coincidence points of the model are compared. The overlap between the effect of roadway and the effect of whole space under the restrictive conditions is analyzed. The results show that the visual stereo image obtained by eliminating abnormal phenomena can better display the response characteristics of the abnormal body of collapse column and be clearer. It can provide reliable theoretical basis for the accuracy of data interpretation of underground electrical prospecting by clearly displaying the distribution range of low resistivity undetermined bodies.