The siamese trackers utilize a dual-branch network to extract object features, significantly improving tracking precision and accuracy. However, they typically rely solely on the object region from the initial frame as a tracking template, neglecting the importance of temporal sequences and spatial changes. This oversight can lead to a decline in tracking performance during sudden changes in the object’s appearance. To address this issue, we propose a novel object tracking framework named TSTrack, which integrates both temporal and spatial features. The core of our design is the fusion of temporal and spatial information to enhance feature representation and robustness. We also implement a dynamic updating mechanism for historical images to reflect the current object state more accurately. On the widely recognized datasets GOT-10k, TrackingNet, and LaSOT, TSTrack achieves average overlap (AO) scores of 77.4%, normalized precision (NP) of 88.6%, and area under curve (AUC) scores of 66.8%. The experimental results demonstrate that TSTrack is advantageous in handling appearance variations, effectively enhancing the overall tracking performance. Furthermore, we conduct a series of ablation experiments to validate the contributions of individual modules to performance improvement.