Early identification of abnormal regions is crucial in preventing the occurrence of underground geotechnical disasters. To meet the high-accuracy detection requirements of underground engineering, this paper proposes a tomography method for abnormal region identification in complex rock-mass structures that utilizes traveltime tomography combined with the damped least-squares method and Gaussian filtering. The proposed method overcomes the limitation of velocity difference in empty region detection and mitigates the impact from isolated velocity mutation in the iteration. Numerical and laboratory experiments were conducted to evaluate the identification accuracy and computational efficiency of forward modeling, including the shortest-path method (SPM), dynamic SPM (DSPM), and fast sweeping method (FSM). The results show that DSPM and FSM can clearly detect abnormal regions in numerical and laboratory experiments. Field experiments were conducted in the Shaanxi Zhènào mine and achieve the reconstruction of the underground roadway distribution. This paper not only realizes the application of abnormal region identification using traveltime tomography but also provides new insight into potential hazards detection in underground geotechnical engineering.