Algorithms for automation of sub tree taxation based on airborne laser scanning and multispectral imagery data

DOI: 10.35595/2414-9179-2025-2-31-21-35

View or download the article (Rus)

About the Authors

Danil A. Bogatyrev

Perm State University,
15, Bukireva str., Perm, 614990, Russia,
E-mail: gis@psu.ru

Artem A. Burdin

Perm State University,
15, Bukireva str., Perm, 614990, Russia,
E-mail: gis@psu.ru

Sergey V. Pyankov

Perm State University,
15, Bukireva str., Perm, 614990, Russia,
E-mail: gis@psu.ru

Abstract

Assessment of forest stands characteristics by remote sensing methods is an important and urgent task in modern forestry, environmental monitoring and natural resource management. In the present study, an assessment of the taxation characteristics of plantings (tree species, height, diameter and growing stock) was carried out based on aerial photography and airborne laser scanning (ALS) data from an unmanned aerial vehicle (UAV) on the example of a mixed forest area in the Berezovsky district of Perm Krai. To validate the results, a ground-based survey of each tree with the definition of the listed taxation characteristics was carried out. The ALS was performed from an altitude of 150 m using the “DJI Matrice 350 UAV”. Segmentation of the ALS data was performed to identify individual trees, resulting in the correct recognition of 962 trees in the test area, which is 10% less than their actual number. Based on the ALS and multispectral survey data, the assessment of taxation characteristics (height, diameter, and wood stocks) was performed using multiple linear regression and random Forest methods. It is shown that the height of trees according to the VLS data is estimated with satisfactory accuracy (RMSE = 1.16 m), while the estimation of trunk diameter (RMSE = 5.40 cm) and wood reserves requires improvement, in particular, the expansion of the training sample. When calculating the wood stock, the model error was 0.348 m3, and the relative error for the entire site was 4.18 %. Also, a classification of tree species based on an orthophotoplane was performed for the test site, which gave a satisfactory result.

Keywords

airborne laser scanning, multispectral imagery, forest stands characteristics, machine learning, random forest

References

  1. Anandakumar R., Nidamanuri R., Krishnan R. Individual Tree Detection from Airborne Laser Scanning Data Based on Supervoxels and Local Convexity. Remote Sensing Applications: Society and Environment, 2019. V. 15. No. 100242. DOI: 10.1016/j.rsase.2019.100242.
  2. Breiman L. Random Forests. Machine Learning, 2001. V. 45. Iss. 1. P. 5–32. DOI: 10.1023/A:1010933404324.
  3. Chen X., Jiang K., Zhu Y., Wang X., Yun T. Individual Tree Crown Segmentation Directly From UAV-borne LiDAR Data Using the PointNet of Deep Learning. Forests, 2021. V. 12. No. 131. DOI: 10.3390/f12020131.
  4. Dalponte M., Coomes D.A. Tree-Centric Mapping of Forest Carbon Density from Airborne Laser Scanning and Hyperspectral Data. Methods in Ecology and Evolution, 2016. V. 7. P. 1236–1245. DOI: 10.1111/2041-210X.12575.
  5. Ho T. The Random Subspace Method for Constructing Decision Forests. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1998. V. 20. No. 8. P. 832–844. DOI: 10.1109/34.709601.
  6. Junttila V., Maltamo M., Kauranne T. Sparse Bayesian Estimation of Forest Stand Characteristics from Airborne Laser Scanning. Forest Science, 2008. V. 54. P. 543–552. DOI: 10.1093/forestscience/54.5.543.
  7. Kauranne T., Pyankov S., Junttila V., Kedrov A., Tarasov A., Kuzmin A., Peuhkurinen J., Villikka M., Vartio V.-M., Sirparanta S. Airborne Laser Scanning Based Forest Inventory: Comparison of Experimental Results for the Perm Region, Russia and Prior Results from Finland. Forests, 2017. V. 8. No. 72. DOI: 10.3390/f8030072.
  8. Kim D.H., Ko C.U., Kim D.G., Kang J.T., Park J.M., Cho H.J. Automated Segmentation of Individual Tree Structures Using Deep Learning over LiDAR Point Cloud Data. Forests, 2023. V. 14 (6). No. 1159. DOI: 10.3390/f14061159.
  9. Li Y., Xie D., Wang Y., Jin S., Zhou K., Zhang Z., Li W., Zhang W., Mu X., Yan G. Individual Tree Segmentation of Airborne and UAV LiDAR Point Clouds Based on The Watershed and Optimized Connection Center Evolution Clustering. Ecology and Evolution, 2023. V. 13. No. e10297. DOI: 10.1002/ece3.10297.
  10. Pu Y., Xu D., Wang H., 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 Sensing, 2023. V. 15. No. 1619. DOI: 10.3390/rs15061619.
  11. Qian C., Yao C., Ma H., Xu J., Wang J. Tree Species Classification Using Airborne LiDAR Data Based on Individual Tree Segmentation and Shape Fitting. Remote Sensing, 2023. V. 15. No. 406. DOI: 10.3390/rs15020406.
  12. Roussel J.R., Auty D., Coops N.C., Tompalski P., Goodbody T.R.H., Sánchez Meador A., Bourdon J.F., De Boissieu F., Achim A. LidR: An R Package for Analysis of Airborne Laser Scanning (ALS) Data. Remote Sensing of Environment, 2021. V. 251. No. 112061. DOI: 10.1016/j.rse.2020.112061.
  13. Sparks A., Smith A. Accuracy of a LiDAR-Based Individual Tree Detection and Attribute Measurement Algorithm Developed to Inform Forest Products Supply Chain and Resource Management. Forests, 2022. V. 13. No. 3. DOI: 10.3390/f13010003.

For citation: Bogatyrev D.A., Burdin A.A., Pyankov S.V. Algorithms for automation of sub tree taxation based on airborne laser scanning and multispectral imagery data. InterCarto. InterGIS. Moscow: MSU, Faculty of Geography, 2025. V. 31. Part 2. P. 21–35. DOI: 10.35595/2414-9179-2025-2-31-21-35 (in Russian)