Abstract: The automatic alignment system is a large system that obtains the beam position by processing the images with particular image processing algorithm, and the position information is used to drive the motors and adjust the beam to the specified position, however, the image quality has a great influence on the accuracy of beam positioning. To avoid the images seriously affected by the noise and beam distortion entering into the automatic alignment process which could generate the misjudgement of alignment results, and to ensure the accuracy of image processing results, this paper tries to describe a simple method to recognize these images, and three main far-field noise evaluation models are established based on the Monte-Carlo simulation. The uncertainty of measurement is regarded as the characteristic parameter to determine whether the image should be processed, and we plot the parameter results to a statistical chart as judgement condition. The experimental results show that, in the case of setting a reasonable threshold, the automatic alignment systems can reject most of the not-good images, and greatly improve the reliability of the alignment system.