To enhance the utilization of pruned leaves from Anji white tea and to achieve rapid detection and quantitative analysis of key quality components in fermented tea, Eurotium cristatum was inoculated to pruned leaves of Anji white tea to prepare Anji golden flower tea, and the changes in the contents of tea polyphenols and tea pigments during the fermentation process were determined. Fourier transform mid-infrared spectroscopy was used to develop quantitative analysis models for the contents of tea polyphenols and tea pigments during the fermentation process. For this purpose, optimization of modeling algorithms, optimization of spectral preprocessing techniques, and feature wavelength selection were carried out. Results showed that random forest regression (RFR) was selected to establish predictive models for the contents of tea polyphenols and tea pigments. The predication model of tea polyphenols obtained using Savitzky-Golay smoothing (SG) combined with iteratively retains informative variables (IRIV) showed the best performance. The optimal spectral preprocessing and variable selection methods were wavelet transform (WT) and successive projections algorithm (SPA) for both theaflavins and thearubigins, while those for theabrownins were first derivative (FD) + multiplicative scatter correction (MSC) and competitive adaptive reweighted sampling, respectively. The models' predictive performance was validated, revealing that the average deviations between predicted and actual values were less than 0.05% for all four tea quality components. Therefore, the quantitative analysis models exhibited good accuracy, enabling rapid detection of the contents of tea polyphenols and tea pigments in Anji golden flower tea. They provide technical support for the high-value utilization of pruned leaves from Anji white tea and the rapid quality monitoring of fermented tea products.