Accurately locating goafs is critical for identifying illegal mining, preventing mining-related geohazards, and facilitating the development and utilization of underground spaces. Conventional methods for locating goafs with InSAR techniques primarily rely on the Probability Integral Model (PIM), which tends to overestimate the ground deformation under subcritical extraction. On the other hand, the number of subcritical extraction working faces significantly rises with mining depth. Under these circumstances, accurately locating subcritical underground goafs using existing methods becomes challenging. To this end, a novel method, which incorporates the improved probability integral model (IPIM) and InSAR technique for locating subcritical goafs, is proposed, named the locating goaf method based on IPIM (LGM-IPIM). Firstly, based on the IPIM, a model between the subcritical goaf parameters and InSAR-derived deformation is built. Then, to reduce the influence of surrounding mining, the goaf azimuth angle is determined with textures and patterns of the InSAR-derived deformation time series. Finally, the genetic algorithm-particle swarm optimization (GA-PSO) is employed to determine the goafs' parameters. The effectiveness of the proposed algorithm has been verified by simulation and real data. The results demonstrate that the proposed LGM-IPIM outperforms conventional methods, presenting the best performance and the highest accuracy. Specifically, compared to the locating goaf method based on PIM (LGM-PIM), the proposed LGM-IPIM improves the location accuracy of goaf boundary points by 28.90% and 86.23% in Areas A and B, respectively.  In addition, the proposed LGM-IPIM has robustness against minor errors within the deformation monitoring and IPIM parameters.