Temperature affects deflection, leading to the inaccurate assessment of a bridge's condition, so it is necessary to separate the deflection effect caused by temperature. This study therefore proposes a joint algorithm that employs singular value decomposition (SVD), wavelet threshold denoising (WTD), and robust local mean decomposition (RLMD) for temperature effect separation. First, the signal underwent initial noise reduction via SVD. WTD was then employed to further eliminate noise and obtain the final noise reduction signal. Finally, RLMD was used to separate the noise reduction signal to identify the daily temperature difference effect, annual temperature difference effect, and long-term deflection of the bridge monitoring signal. The results revealed that the effect of SVD-WTD joint noise reduction is superior to using SVD or WTD methods alone, and the accuracy of separating the various temperature effects is significantly improved. The daily temperature difference effect separated from the measured data is continuous in time, and the daily and annual temperature difference effects at the corresponding measurement points are spatially correlated. The deflection components separated from the simulated signals and measured data have periodicity and similar change trends, verifying that the proposed method can effectively separate the temperature effect components in the bridge deflection monitoring signals.