Owing to the presence of noise and the Doppler background, accurate saturated absorption (SA) peak automatic identification technology poses a significant challenge for laser frequency tuning and locking. To address this issue, a novel peak identification algorithm for the SA spectrum is proposed. First, a Gaussian filter based on a Gaussian continuous wavelet transform is proposed to mitigate the spectral high-frequency noise. Subsequently, a hybrid method combining a first-order Gaussian continuous wavelet transformation and adaptive threshold judgment was designed for multi-peak boundary segmentation. Finally, we obtained the target peak and its sweeping voltage based on an adaptive nonlinear fitting algorithm, which was almost unaffected by the peak asymmetry caused by the Doppler background.