It is common for large enterprises or organizations to maintain repositories of process models. A large number of process model variants may exist in these repositories due to the differences of modeling objective or scenario, customization or tailoring reference models, and model updating modification. This paper focuses on the study of identifying commonalities and differences between process variants, i.e. construction of matching relations between process variants automatically. To support the discovery of complex correspondences and ensure the effectiveness and efficiency of matching results, we propose a matching technique based on the traversal of process structure tree and present a process similarity measuring method based on tree-edit distance. The experimental evaluation based on real-world process model collection shows that an effective precision and recall is achieved.