Fig. 1. Flow chart of method

Fig. 2. Schematics of constructing shortest path. (a) Constructing shortest path with point spacing as weight; (b) constructing shortest path with square of point spacing as weight

Fig. 3. Constructed undirected graph structure and trimmed graphs. (a) Constructed undirected graph; (b) result 　of graph segmentation

Fig. 4. Skeletons of branches and trunks extracted by shortest path algorithm. (a) Original point cloud of branches and trunks; (b) skeleton of branches and trunks extracted with distance between points as weight; (c) skeleton of branches and trunks extracted with square of distance between points as weight

Fig. 5. Point cloud of tree branches and trunks extracted by shortest path analysis algorithm. (a) Original point cloud of branches and trunks; (b) extracted point cloud of branches and trunks

Fig. 6. Point cloud of tree branches and trunks extracted by graph-based segmentation algorithm. (a) Original point cloud of branches and trunks; (b) extracted point cloud of branches and trunks

Fig. 7. Merging flow chart of tree point cloud separation results

Fig. 8. Single tree point clouds with different average point spacings. (a) Tree 1; (b) Tree 2; (c) Tree 3

Fig. 9. Single-tree point clouds for different data quality or different tree species. (a) Tree 4; (b) Tree 5; (c) Tree 6; (d) Tree 7; (e) Tree 8; (f) Tree 9; (g) Tree 10; (h) Tree 11; (i) Tree 12; (j) Tree 13; (k) Tree 14; (l) Tree 15; (m) Tree 　16; (n) Tree 17; (o) Tree 18; (p) Tree 19

Fig. 10. Box-and-whisker plot

Fig. 11. Separation results for branches and leaves of Tree 3 and reference values in open source data set. (a) Separation result for branches and leaves by proposed method; (b) reference value for branch and leaf separation in open source data set

Fig. 12. Separation results for branches and leaves by proposed method . (a) Tree 1; (b) Tree 2; (c) Tree 3

Fig. 13. Separation results for branches and leaves by TLSeparation method. (a) Tree 1; (b) Tree 2; (c) Tree 3

Fig. 14. Separation results for branches and leaves by LeWos method. (a) Tree 1; (b) Tree 2; (c) Tree 3

Table 1. Tree point cloud data

Table 2. Pseudo-randomly selected input parameter values for verification testing

Table 3. Separation result for branches and leaves of each tree

Table 4. Separation results for branches and leaves from different trees

Table 5. Evaluation indexes for TLSeparation method in separation of branches and leaves from 19 trees

Table 6. Evaluation indexes for LeWos method in separation of branches and leaves from 19 trees

Table 7. Evaluation indexes for proposed method in separation of branches and leaves from 19 trees