Eye movement information has been proven to be a new type of data that can integrate human expert knowledge into intelligent models. However, currently, there is a lack of the eye movement datasets for cross-view matching tasks, especially for the task of street-to-aerial images cross-view matching. In this paper, we introduce an eye movement dataset based on this cross-view matching task, named Eye-CVACT. Five college students with relevant expertise participate in the experiment. We analyze the behavioral data of the subjects and apply this dataset to the visual saliency prediction driven by this task. The results verify the effectiveness and scientific value of this dataset. The Eye-CVACT dataset not only facilitates research on hybrid intelligent models for cross-view matching, but also contribute to the research on the interpretability of intelligent models.