Abstract: Accurate identification of concealed geological anomalies within longwall panels is essential for safe and efficient deep coal mining. A high-resolution seismic full-waveform tomography method based on a high-order staggered-grid finite-difference scheme and the adjoint-state method was developed. A convolution-based objective function independent of the source wavelet was constructed, and iterative inversion was performed using a multiscale strategy and wavefield reconstruction. Two-dimensional elastic-wave numerical simulations were conducted for the floor of a longwall panel, with collapse columns and fault fracture zones considered as representative geological anomalies. The results show that the proposed inversion method converged effectively and accurately reconstructed the locations, geometries, and velocity structures of the anomalous bodies, demonstrating the reliability of the method. Field application further showed that the method is not constrained by coal-seam continuity and provides greater investigation depth and higher resolution. It can effectively identify concealed geological anomalies such as small faults and collapse columns, providing reliable technical support for safe deep coal mining.