Enhancing EOM with density-adaptive downsampling for individual tree segmentation in density-heterogeneous forest point clouds

Abstract

Addressing the insufficient segmentation accuracy caused by forest point cloud heterogeneity in existing extreme offset networks for individual tree segmentation, this study proposes a density-adaptive downsampling model with adaptive point cloud density sampling, multi-scale dynamic feature fusion, sample relationship enhancement. First, to resolve key feature points omission from uneven density distribution, the adaptive point cloud density sampling method enhances sparse region sampling probability through local density-weighted scoring. Second, multi-scale dynamic feature fusion module adaptively adjusts neighborhood scales based on local density and integrates cross-scale features. Finally, to clarify tree boundaries, sample relationship enhancement module with external learnable memory units captures global contextual information. For evaluating the method’s accuracy and generalizability across varying forest structures and point cloud densities, experiments are conducted on two low-density coniferous forest plots (U1, U2) in Washington’s Blue Ridge area (USA) and two high-density mixed forest plots (G1, G2) in Bretten (Germany). Results demonstrate average precision, recall, and F1-score of 0.92, 0.92, and 0.92, respectively, outperforming the baseline EOM (p = 0.91, r = 0.88, F-score = 0.90) with improvements of 1%, 4%, and 2% in precision, recall, and F-score.

Keywords

Individual tree extraction / Extreme offset deep learning / Adaptive point cloud density sampling / Multi-scale dynamic feature fusion / Sample relationship enhancement

References

[1] Brodić N, Cvijetinović Ž, Milenković M, Kovačević J, Stančić N, Mitrović M, Mihajlović D. Refinement of individual tree detection results obtained from airborne laser scanning data for a mixed natural forest. Remote Sens, 2022, 14(21. ArticleID: 5345

[2] Chen QJ, Wang X, Hang MR, Li JY. Research on the improvement of single tree segmentation algorithm based on airborne LiDAR point cloud. Open Geosci, 2021, 13(1): 705-716.

[3] Fu HP, Li H, Dong YQ, Xu F, Chen FX. Segmenting individual tree from TLS point clouds using improved DBSCAN. Forests, 2022, 13(4): 566.

[4] Guo MH, Liu ZN, Mu TJ, Hu SM. Beyond self-attention: external attention using two linear layers for visual tasks. IEEE Trans Pattern Anal Mach Intell, 2023, 45(5): 5436-5447.

[5] Guo ML, Zhang HQ, Tan JW, Liu Y, Chen SH, Lei H, Shi YK. Spatiotemporal prediction of forest litterfall in China by using multi-source data and Transformer-CatBoost model. J for Res, 2025, 37(1): 24.

[6] Hui ZY, Jin SG, Li DJ, Ziggah YY, Liu B. Individual tree extraction from terrestrial LiDAR point clouds based on transfer learning and Gaussian mixture model separation. Remote Sens, 2021, 13(2): 223.

[7] Jing WP, Xu HM, Zou WT, Zhang WJ, Li C, Gu JT. Advancing Acer phenology monitoring: fine-grained identification and analysis by deep learning RESformer. J for Res, 2025, 36(1): 54.

[8] Kong M (2021) Single tree extraction of airborne forest point cloud based on deep learning of candidate regions. Master’s thesis, China University of Geosciences, Beijing, China (in Chinese)

[9] Li WK, Guo QH, Jakubowski MK, Kelly M. A new method for segmenting individual trees from the lidar point cloud. Photogramm Eng Remote Sens, 2012, 78(1): 75-84.

[10] Lv AK (2021) Research on individual tree segmentation method of coniferous forest based on airborne LiDAR. Master’s thesis, Northeast Forestry University, Harbin, China (in Chinese)

[11] Moenning C, Dodgson NA (2003) Fast marching farthest point sampling. Technical Report UCAM-CL-TR-562, University of Cambridge Computer Laboratory, Cambridge. https://doi.org/10.48456/tr-562

[12] Pu YH, Xu DD, Wang HB, Li X, Xu X. A new strategy for individual tree detection and segmentation from leaf-on and leaf-off UAV-LiDAR point clouds based on automatic detection of seed points. Remote Sens, 2023, 15(6): 1619.

[13] Qi CR, Yi L, Su H, Guibas LJ (2017) PointNet++: deep hierarchical feature learning on point sets in a metric space. Preprint at https://arxiv.org/abs/1706.02413

[14] Rocha KD, Silva CA, Cosenza DN, Mohan M, Klauberg C, Schlickmann MB, Xia JY, Leite RV, de Almeida DRA, Atkins JW, Cardil A, Rowell E, Parsons R, Sánchez-López N, Prichard SJ, Hudak AT. Crown-level structure and fuel load characterization from airborne and terrestrial laser scanning in a longleaf pine (Pinus palustris Mill.) forest ecosystem. Remote Sens, 2023, 15(4. ArticleID: 1002

[15] Tian S, Long AY. Point cloud classification method based on graph convolution and multilayer feature fusion. Laser Optoelectron Prog, 2023, 60(14): 1415003. (in Chinese).

[16] Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, Kaiser L, Polosukhin I (2017) Attention is all you need. Preprint at https://arxiv.org/abs/1706.03762

[17] Wang FY, Bryson M. Tree segmentation and parameter measurement from point clouds using deep and handcrafted features. Remote Sens, 2023, 15(4. ArticleID: 1086

[18] Wang XH, Zhang YZ, Luo Z. Combining trunk detection with canopy segmentation to delineate single deciduous trees using airborne LiDAR data. IEEE Access, 2020, 8: 99783-99796.

[19] Wang LL, Zhang HQ, Fu RR, Lei KX, Liu Y, Yang TD, Zhang J, Ge XN. MTSCFNet: a novel framework for improving tree species classification in a subtropical forest using RGB, LiDAR-derived, and GF-2 data. J for Res, 2025, 37(1): 22.

[20] Weiser H, Schäfer J, Winiwarter L, Krašovec N, Fassnacht FE, Höfle B. Individual tree point clouds and tree measurements from multi-platform laser scanning in German forests. Earth Syst Sci Data, 2022, 14(7): 2989-3012.

[21] Wu J, Cui Y, Zhao XM, Chen RX, Xu G. SSA-point net++: a space self-attention CNN for the semantic segmentation of 3D point cloud. J Comput Aided des Comput Graph, 2022, 34(3): 437-448. (in Chinese).

[22] Xia K, Li C, Yang YH, Deng SS, Feng HL. Study on single-tree extraction method for complex RGB point cloud scenes. Remote Sens, 2023, 15(10): 2644.

[23] Yan WQ, Guan HY, Cao L, Yu YT, Li C, Lu JY. A self-adaptive mean shift tree-segmentation method using UAV LiDAR data. Remote Sens, 2020, 12(3): 515.

[24] Yun T, Jiang K, Hou H, An F, Chen BQ, Jiang AN, Li WZ, Xue LF. Rubber tree crown segmentation and property retrieval using ground-based mobile LiDAR after natural disturbances. Remote Sens, 2019, 11(8. ArticleID: 903

[25] Yun T, Jiang K, Li GC, Eichhorn MP, Fan JC, Liu FZ, Chen BQ, An F, Cao L. Individual tree crown segmentation from airborne LiDAR data using a novel Gaussian filter and energy function minimization-based approach. Remote Sens Environ, 2021, 256. ArticleID: 112307

[26] Zhang YY, Wu HF, Mei XJ, Xian JF, Wang WJ, Zhang QN, Liang LN. Two-phase robust target localization in ocean sensor networks using received signal strength measurements. Sensors Basel, 2021, 21(5. ArticleID: 1724

[27] Zhang HC, Zhang HQ, Xu KQ, Li YQ, Wang LL, Liu R, Qiu HQ, Yu LH. A novel framework for stratified-coupled BLS tree trunk detection and DBH estimation in forests (BSTDF) using deep learning and optimization adaptive algorithm. Remote Sens, 2023, 15(14): 3480.

[28] Zhang YZ, Liu HT, Liu XY, Yu HL. Towards intricate stand structure: a novel individual tree segmentation method for ALS point cloud based on extreme offset deep learning. Appl Sci, 2023, 13(11): 6853.

[29] Zhao HS, Jiang L, Jia JY, Torr P, Koltun V (2022a) Point transformer. In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, QC, Canada. IEEE, pp 16239–16248.

[30] Zhao L, Xu SY, Liu LM, Ming DL, Tao WB. SVASeg: sparse voxel-based attention for 3D LiDAR point cloud semantic segmentation. Remote Sens, 2022, 14(18): 4471.

[31] Zhao H, Jiang L, Jia J, Torr PHS, Koltun V (2021) Point transformer. In: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp 16259–16268.

[32] Zhu LP, Wang BY, Tian GY, Wang WJ, Li CY. Towards point cloud completion: point rank sampling and cross-cascade graph CNN. Neurocomputing, 2021, 461: 1-16.

RIGHTS & PERMISSIONS

The Author(s)