Abstract: The extensive application of deep learning technology in the field of computer vision has promoted the rapid development of technologies such as intelligent security, intelligent transportation, and autonomous driving. Among them, the automatic detection and tracking of pedestrians is the prerequisite for realizing intelligence and automation. But nowadays, accurate detection and tracking of pedestrians in low-light scenes at night still faces great difficulties. In order to solve the problems of difficult target feature extraction and unstable tracking in low-light scenes at night, this paper proposes a multi-target pedestrian detection and tracking algorithm based on the autoencoder structure and improved Bytetrack. In the detection phase, this paper builds a multi-task auto-encoding transformation model framework based on YOLOX, considers the physical noise model and image signal processing (ISP) process in a self-supervised manner, and learns the intrinsic features by encoding and decoding the real illumination degradation transformation process. visual structure, and implement object detection tasks by decoding bounding box coordinates and classes based on this representation. In order to suppress the interference of background noise, the adaptive feature fusion module ASFF is introduced in the target decoder neck network. In the tracking phase, the Bytetrack algorithm is improved, and the appearance embedded information extracted based on the Transformer re-identification network and the motion information obtained by the NSA Kalman filter are used to complete the data association through an adaptive weighting method, and the Byte twice matching algorithm is used to complete the data association at night. Pedestrian tracking. The generalization ability of the detection model was tested on the self-built night pedestrian detection data set, and mAP@0.5 reached 94.9%. The results show that the degradation transformation process in this paper meets the realistic conditions and has good generalization ability. Finally, the multi-target tracking performance is verified through the night pedestrian tracking data set. The experimental results show that the MOTA of the night low-light pedestrian multi-target tracking algorithm proposed in this paper is 89.55%, IDF1 is 88.34%, and IDs are 15. Compared with the baseline method Bytetrack, MOTA improves by 10.72%, IDF1 improves by 6.19%, and IDs are reduced by 50%. The results show that the multi-target tracking algorithm based on the autoencoding structure and improved Bytetrack proposed in this article can effectively solve the problem of difficult pedestrian tracking in low-light scenes at night.