Abstract: Retrieving 3D tree model using generic retrieval methods often encounter problems such as poor accuracy, semantic mismatch, and low efficiency. To overcome these problems and improve accuracy and efficiency of 3D tree model's retrieval, a progressive retrieval method for 3D tree model based on local view is proposed. This method adopts hierarchy local view to describe tree model, and highlights the important features of tree model(e. g. crown contour, branches skeleton) in a simple and efficient form. Then, tree semantics is introduced to extract view features, which makes view feature be distinguished more easily for different tree models. At last, view feature is adopted in the multi-stage retrieval process that achieves a coarse-to-fine retrieval for 3D tree models, and dynamic index is used to speed up the retrieval process. Experimental results show that this method achieves better performance in tree model's retrieval both for general model benchmark and tree model benchmark, thus it make 3D tree model resource be reused in an easy and efficient way.