Abstract: This paper proposes an algorithm for extracting skeleton network of areal object and organizing it into hierarchical structure based on multivariate analysis. We first extract the primary skeleton network by TIN of the areal object. Then an evaluation model, which is a function incorporating length, angle and position of edges, is set up to get a weight value for each edge of the skeleton network. Finally, a dynamic priming process considering the weights is employed to produce the hierarchical structure.