The on-site measured data of short-wave irregularities often exhibit non-stationary on urban rail transit track surfaces. This paper employs the wavelet method with multi-resolution and adaptivity is used to study the local characteristics, defect identification and track irregularity power spectrum of non-stationary track surface shortwave irregularity. Daubechies10 is selected as the wavelet basis function to identify the waveform characteristics of rail surface shortwave irregularity. Using the Daubechies10 wavelet, the signal is decomposed into 10 layers, and the power spectrum analysis of the decomposed detail signal and the approximation signal is carried out to obtain the contribution of different waveforms to the power spectrum. The wavelet power spectrum of short-wave irregularity of the rail surface in the joint area of the straight section and the curve section of the two track types of the elevated line and the underground line are calculated.The results demonstrate that wavelet basis functions can effectively identify local features of short-wave irregularity waveforms. Through the approximation and detail signals obtained from wavelet decomposition, significant irregularities at wavelengths of 20.298 m and 115.98 m were identified. Power spectrum analysis of these decomposed signals revealed the contributions of different waveforms to the power spectrum across various wavelength ranges. The results indicate that the wavelet analysis method can effectively process non-stationary data, providing a robust approach to comprehensively analyze the characteristics of short-wave irregularities on urban rail transit track surfaces.