In the field measurement, different signal to noise ratios (SNR) have a great impact on the experimental results with the same instrument and data processing methods. By analyzing the noise model measured in the field, the same noise is added to the spot image simulated to change the SNR while the signal intensity is kept unchanged. When the SNR is less than 5, the relative deviation increased significantly. Then different methods are used to subtract the background noise under the same SNR, and it is found that the Gaussian noise has little effect on the centroid position after subtracting the average background value. The results show that the overall performance of the subtraction of 1 times average background and 0.6 times background fluctuation standard deviation is found to be the best.