The application of vehicle vision detection and tracking technology in the domain of vehicle transportation management within container terminals holds significant value. However, challenges persist in accurately detecting and identifying targets, particularly in complex scenes characterized by occlusion. These challenges, including misdetection, omission, and identity switching, underscore the need for further research and development in multi-scale feature extraction and target detection algorithms. To this end, this paper proposes a multi-target vehicle visual detection and tracking method based on the composition of an improved YOLOv8 detector (YOLO-FAS) and an improved ByteTrack algorithm. In the context of vehicle detection, the YOLO-FAS algorithm demonstrates notable enhancements in feature retention during the downsampling process. This enhancement is achieved through the integration of an adaptive dual-pooling downsampling module (ADown), which facilitates the optimization of multi-scale feature fusion through the incorporation of a dense weight residual module (DWR). Furthermore, the adoption of a partial convolution module (CSPPC) contributes to the lightweighting and acceleration of the model, thereby enhancing the efficiency of target detection in complex scenarios. In the context of vehicle tracking, the extended Kalman filter is employed to enhance the precision of target motion prediction. Additionally, the target association robustness under occlusion and similar appearances is notably augmented by integrating the affinity network (DAN) module. The experimental results on the container terminal scenario dataset demonstrate the efficacy of the algorithm proposed in this paper, exhibiting superior performance in terms of detection and tracking. The detection accuracy metric mAP 0.5 has been enhanced to 99.5%, and the tracking performance metrics MOTA and IDF1 have been improved to 96.4% and 88.1%, respectively. The efficacy of the enhanced algorithm is further substantiated by comparison experiments with mainstream detection and tracking methods. YOLOv8 · ByteTrack · Container terminal.