Abstract: This article used Guiyang Metro Line 3 under construction as an example to study the application of cross hole seismic computed tomography (CT) technology in the exploration process of Karst complex areas, thereby revealing the characteristics of Karst development in the study area. Through comparative analysis of various algorithms, it was determined that compared to the Conjugate Gradient Least Squares (CGLS) algorithm, the Sparse Linear Equations and Least Square Problems (LSQR) algorithm is more sensitive and accurate in identifying low-speed anomalies, and has better convergence and reliability. Drilling verification shows that using the LSQR algorithm and bending ray tracing method for seismic tomography inversion can achieve meter level accuracy in cross hole seismic CT detection, obtain high-quality velocity images, and identify the depth, scale, and extension of Karst between boreholes, neutralizing the limitations of drilling. This indicates that seismic CT technology based on the LSQR algorithm is practical and feasible for detecting Karst.