Abstract: At present, most of the techniques for inverting coal thickness through channel wave dispersion are based on the theoretical dispersion curve to select a single frequency for velocity CT tomography imaging, and the accuracy of the inverted coal thickness in the working face is not high. To improve the prediction accuracy of coal thickness in coal seam working faces, the channel wave dispersion curve of the multi-layer elastic medium model is calculated, and the effectiveness of the three-dimensional dispersion inversion method was verified through forward simulation. The dispersion curve of the Love channel wave in the actual coal seam was extracted by using the generalized S-transform. Tomographic imaging of the working face was performed by setting the frequency range and step size, and the velocity CT imaging maps of the working face at different frequencies were obtained. The working face is divided into several sections. The dispersion curves of each section are calculated based on the results of tomographic imaging. The dispersion curves of each section are inverted, and the velocity distribution and coal thickness distribution of the entire working face are obtained by integrating the inversion results. The dispersion inversion algorithm selects nonlinear global optimization algorithms, such as genetic algorithm, pattern search algorithm, particle swarm optimization algorithm and simulated annealing algorithm, etc. This type of algorithm is more likely to find the global optimal solution and thereby improve the accuracy of coal thickness inversion. The three-dimensional dispersion inversion technology of channel waves in the coal mining face of a certain coal mine is selected for verification. The average coal thickness of this working face was 3.7 m. The double-channel transmission method was adopted for channel wave seismic exploration. The generalized S-transform was used to extract each dispersion curve of the Love channel wave, and the velocity was extracted once every 10 Hz within the frequency band of 100−250 Hz. The CT imaging results of the working face velocity at 16 frequencies were obtained by using tomography. Then, each dispersion curve of the working face was calculated based on the tomography results, and the entire dispersion curve was inverted by using the genetic algorithm. The results show that the three-dimensional dispersion inversion of channel waves improves the accuracy of coal thickness prediction in coal seam working faces. Compared with the coal thickness exposed in the roadway, the average fitting error is only 0.22 m. Therefore, the three-dimensional dispersion inversion imaging technology of channel waves has a relatively high accuracy in the prediction of coal thickness in the working face.