To address the low detection accuracy of existing laser-based rail corrugation inspection methods, a rail corrugation signal processing approach based on grey wolf optimization-variational mode decomposition (GWO-VMD) is proposed. In this method, multi-scale permutation entropy (MPE) and variance contribution rate (VCR) are combined to construct a dual-threshold criterion for accurate identification of noise components. Subsequently, non-local means (NLM) filtering is applied to denoise the identified noise components. The denoised components are then reconstructed together with the effective components, and the moving peak-to-peak value (PPR), moving root mean square of corrugation depth amplitude (RMS), and exceedance rate are calculated to achieve quantitative identification of the corrugation wavelength, location, and wear severity.Using measured data from the Xiangyu Railway as a case study, the proposed algorithm is compared with several representative methods, including particle swarm optimization-variational mode decomposition combined with non-local means (PSO-VMD-NLM), variational mode decomposition combined with non-local means (VMD-NLM), ensemble empirical mode decomposition combined with non-local means (EEMD-NLM), grey wolf optimization combined with variational mode decomposition combined with wavelet thresholding (GWO-VMD-WT), complete ensemble empirical mode decomposition with adaptive noise combined with wavelet thresholding (CEEMDAN-WT), complete ensemble empirical mode decomposition with adaptive noise combined with non-local means (CEEMDAN-NLM), and the traditional wavelet thresholding (WT) denoising method. Experimental results demonstrate that, compared with the above methods, the proposed algorithm improves the signal-to-noise ratio (SNR) by 3.9%, 32.4%, 68.4%, 8.7%, 46.6%, 23.6%, and 89.6%, respectively, and reduces the root mean square error (RMSE) by 15.5%, 74.4%, 85.9%, 13.1%, 80.5%, 68.8%, and 88.9%, respectively.Furthermore, generalization capability is validated using measured data from multiple railway lines. The results indicate that the proposed algorithm maintains stable denoising performance and identification accuracy under new line conditions. It enables multi-band separation and multi-feature quantitative analysis of rail corrugation under complex noise conditions, providing reliable technical support for track condition assessment and precision grinding maintenance.