Abstract: Tree modeling technology based on point cloud is an important method for obtaining threedimensional model of trees. The tree branch point cloud data measured by 3D laser scanner were applied as data source, and the point cloud data of the tree branches along the tree height were divided into different layers to establish the hierarchical model of the trees. The contour of the point cloud data at different tree height was extracted by using convex hull algorithm, and the Delaunay triangulation between the adjacent contours was constructed by triangulation growth algorithm. The Delaunay triangulation model of the whole tree trunk was built by integrating the Delaunay triangulations in different parts of the tree. The realistic examples verified that the contour model of the branches extracted by this method was consistent with common contour model, and there would be no edge crosscutting problems between two adjacent contours. The data quantity requested for building Delaunay triangulation model with point cloud data of the tree branches was much less than the building with other ordinary software, and the model building effect was much better. It was showed that the error of the model parameters extracted by ray collision detection technique was less than 5% compared with the actual measurement value, which could meet the demand of forestry tree measurement.