Near-surface geological defects pose a serious threat to human life and infrastructure. Hence, the exploration of geological hazards is essential. Currently, there are various geological hazard exploration methods; however, those require improvements in terms of economic feasibility, convenience, and lateral resolution. To address this, this study examined an extraction method to determine spatial autocorrelation velocity dispersion curves for application in near-surface exploration. The influence of the spatial autocorrelation coefficient calculation error on the velocity dispersion curves was determined using the theoretical calculation formula. Advantageous frequency bands with different radii were obtained and the extraction method of the velocity dispersion curve for multi-radius advantageous frequency bands was summarized. Finally, the method was applied to explore tunnel damage in a representative section of a railway tunnel. Through a comparative analysis of the results of multi-channel surface wave analysis and drilling methods, this method was observed to accurately identify abnormal areas of geological structures. Compared with traditional processing methods, the method described herein has a higher lateral resolution and can effectively suppress the influence of interference factors on the calculation results, thereby improving the accuracy of the velocity dispersion curves. The results of this study provide a reference for the application of spatial autocorrelation methods in the investigation of urban near-surface geological defects.