Abstract: The water level of leachate will affect the stability of landfill and have the risk of leakage and pollution. When the leachate is stored on the HDPE impermeable membrane, the extreme differentiation characteristics of the resistivity characteristics of the two and the boundary effect and other factors make the least squares and other traditional geophysical inversion methods unable to accurately invert the actual resistivity distribution, and then according to the resistivity the difference feature locates the height of the leachate water level above the HDPE membrane. In order to accurately describe the fine distribution of the local resistivity of the leachate-HDPE membrane inside the garbage dump, The traditional high density electrical method (ERT) device is improved, and a detection device (C-ERT) is proposed, and the resistivity inversion model algorithm of BP neural network is adopted. The method is verified by COMSOL theoretical model and field data collected from a domestic waste landfill in Jiangxi Province, and compared with the least square algorithm (LS). The results show that the BP algorithm based on C-ERT can effectively identify the leachate area above HDPE membrane, and the recognition accuracy is about 83.2%, while LS inversion algorithm can not identify the leachate area.