Abstract: Accurate identification of concealed karst is fundamental to underground space development and geological engineering safety. However, in regions characterized by low-resistivity clay overburden and complex hydrogeological conditions, traditional Electrical Resistivity Tomography (ERT) is significantly hindered by the "shielding effect" of highly conductive media. This effect obscures the electrical response of deep-seated targets, thereby limiting detection resolution and accuracy. To address the issues of blurred imaging and imprecise boundary characterization under complex overburden, this study aims to construct a constrained inversion framework that integrates multi-source information to enhance the accurate identification of concealed karst zones. The attenuation patterns of electrical signals from caves under low-resistivity covers were quantitatively analyzed through finite element method forward modeling. Subsequently, a multi-source constrained ERT inversion algorithm, incorporating seismic reflection interfaces as structural constraints and borehole prior data as physical constraints, was developed based on the pyGIMLi platform. This approach directly incorporated the seismic geometric framework and borehole-based physical calibration into the inversion objective function and was validated using a representative karst-prone field site. The results indicate that, compared with independent inversion, the multi-source constrained inversion achieves lower data fitting errors across two representative survey lines and higher identification accuracy for concealed karst. In comparison with borehole data, the anomaly boundaries obtained from the constrained inversion are more focused, effectively correcting the longitudinal smearing artifacts common in traditional independent inversion. Notably, the depth identification accuracy of the karst bottom boundaries is improved by 5 to 7 meters at EC3*. Furthermore, the method corrects the misidentification of isolated anomalies (formerly designated as EC4 and EC5), revealing the lateral connectivity features of the karst zone. The study indicates that integrating structural and physical information into the inversion process significantly improves the accuracy of concealed karst resistivity imaging. These research findings provide a reliable technical solution for accurate karst detection and hazard mitigation under complex overburden, offering broad application value for ensuring the safety of large-scale infrastructure construction.