【Objective】 The tree measurement in a virtual environment make plant modeling become a continuous concern. The objective of this study is to develop a new methodology for the reconstruction of 3D surface models of individual trees based on TLS point clouds that are expected to obtain precise tree measurement factors. 【Method】 A k-d (k-dimensional) tree was used to build adjacent relations between branches after cloud point preprocessing, and the root of the subgraph was calculated using the Dijkstra algorithm. Furthermore, when the shortest path is established, the critical path is detected by the detection radius. After the branch skeletons were calculated, a Bezier curve was used to smooth the initial radius, and the smoothed skeletons of the branch and trunk were connected. In this situation, the smoothed radius and cylinder fitting were applied to retain the details of the branches if the cloud point density was too small. 【Result】3D surface models of branches and trunks for three larch trees were built, and the tree architecture was reconstructed. The matching of the point cloud and surface models demonstrated good behavior. Subsequently, the small branches were reconstructed instead of simulation, and the model can display a good effect of first-level branches, and bigger second-level branches can also be shown on the screen. The time consumed by this algorithm was extremely fast, and depended on the complexity and connection between branches. 【Conclusion】 The methodology based on critical path detection is capable of reconstruction of 3D tree surface modeling that can be applied for precise measurement of individual trees.