Abstract: Aiming at the low-detection precision of traditional geophysical prospecting inversion methods for unfavorable geological conditions such as karst caves in the early detection stage of infrastructure projects, this study proposed a cross-hole resistivity detection and inversion method of karst caves based on the squirrel search algorithm to improve the performance of the traditional Tikhonov regularization-based sensitivity iteration method, which is sensitive to initial values and noise and easy to fall into local optimization. The detection results obtained using different intelligent search algorithms and sensitivity iteration methods were compared and analyzed using three numerical examples of small, large, and beaded karst caves. Moreover, an indoor physical model was also built to validate the proposed method. The results show that the inversion method based on the squirrel search algorithm has a high convergence speed and precision and can significantly improve the detection precision of karst caves using the cross-hole resistivity method.