Abstract: This paper proposed a railway foreign object intrusion limit tracking method based on background perception correlation filters to address the safety issues of train operation caused by frequent foreign object intrusion limit in railways. The paper utilized the HOG (Histogram of Oriented Gradient) feature extraction to extract the characteristics of railway intrusion foreign objects, and combined it with the clipping matrix to increase the negative samples of the actual background in the video frame, used the ADMM (Alternating Direction Method of Multipliers) to train background perception related filters model, reduce computational complexity and improving the accuracy of tracking intruded foreign objects while ensuring tracking speed, thus adapted to the situations of target loss and tracking frame drift caused by deformation, rapid movement or weather of intruded foreign objects in railway environments. The experimental results show that the tracking accuracy and AUC (Area Under Curve) value of this method for railway intrusion foreign objects reach 93% and 71.9%, respectively, which are higher than algorithms such as SRDCF, KCF, ASLA, and CSK, and have better accuracy.