Real-time and precise passenger trajectory tracking is a key technology for achieving intelligent safety management in subway stations. However, subway stations commonly suffer from low-light conditions and frequent passenger occlusions, which often compromise the tracking accuracy of existing mainstream multi-object tracking (MOT) techniques of computer vision in practical applications of subway management. To address this issue, this paper proposes a passenger trajectory tracking method based on low-light image enhancement, aiming to improve tracking performance in low-light subway station environments and frequent passenger occlusions. First, an advanced Retinexformer technique is introduced to restore details and enhance image quality in low-light surveillance videos. Subsequently, two MOT algorithms, i.e., the observation-centric SORT (OC-SORT) and the feature-sparse SparseTrack, are employed for passenger trajectory tracking. Experimental results demonstrate that after image enhancement, the OC-SORT model achieves 73% HOTA and 96% MOTA in low-light conditions, while the SparseTrack model achieves 76% HOTA and 98% MOTA.