Accurate image-based tracking is essential for displacement measurement in high-speed mechanical testing, particularly when conventional contact-based or single-point methods are unable to capture the full response of a specimen. However, high-speed imaging is often affected by insufficient illumination, partial marker loss, and target deformation, which can significantly reduce tracking accuracy. To address these challenges, this study integrates the open-source Swin transformer (shifted-window transformer) into the commercially available YOLOv8 object detection framework developed by Ultralytics and proposes an improved method for marker detection and displacement measurement in high-speed photogrammetry. The proposed model combines the hierarchical feature extraction capability of the Swin transformer with the multiscale feature fusion strategy of YOLOv8, thereby enhancing the detection performance for small and degraded targets under challenging imaging conditions. Moreover, the structural displacement is calculated by applying a weighted averaging scheme to the spatial coordinates of the identified markers. Experimental findings show that the model augmented with the Swin transformer delivers improved performance in both marker detection and displacement evaluation tasks. The proposed method enables high-accuracy displacement measurement under a wide range of lighting conditions and therefore shows strong potential for practical applications in high-speed mechanical testing.