Abstract
Temporal image data association is essential for visual object tracking tasks. This association task is typically stated as a process of connecting signals from the same object at different times along the time axis. Temporal data association is usually performed before state estimation. The accuracy of data association processing results is fundamental to guaranteeing the correctness of all subsequent procedures. This paper proposes an efficient approach for temporal data association focused on obtaining accurate data association processing results in Siamese network framework. Siamese network has recently achieved strong power in visual object tracking owing to its balanced accuracy and speed. Based on data association processing and multi-tracker collaboration, our algorithm achieves high accuracy and strong robustness, which outperforms several state-of-the-art trackers, including standard Siamese trackers.