Abstract: Objective Road network generalization and analysis have been hot issues in the geographical informa-tion science.The stroke in a road network is defined as a set of one or more road segment in a non-branching,connected chain,based on the good continuity principle.The stroke plays an importantrole in road network generalization,analysis,pattern recognition and schematic map generation.Ex-isting studies focus on the stroke generation algorithms.However,the formalization of stroke genera-tion at the conceptual level is absent.This paper first formalizes the stroke generation as clusteringproblem,and then presents a hierarchical clustering based stroke generation algorithm.Time com-plexity and some properties of the proposed algorithm are analyzed in detail.Finally,the algorithm isverified using the Shenzhen road network at the 1∶50 000scale.