Automated Segmentation of Individual Tree Structures Using Deep Learning over LiDAR Point Cloud Data
Deep learning techniques have been widely applied to classify tree species and segment tree structures. However, most recent studies have focused on the canopy and trunk segmentation, neglecting the branch segmentation. In this study, we proposed a new approach involving the use of the PointNet++ mo...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-06-01
|
| Series: | Forests |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1999-4907/14/6/1159 |
| _version_ | 1797594789487575040 |
|---|---|
| author | Dong-Hyeon Kim Chi-Ung Ko Dong-Geun Kim Jin-Taek Kang Jeong-Mook Park Hyung-Ju Cho |
| author_facet | Dong-Hyeon Kim Chi-Ung Ko Dong-Geun Kim Jin-Taek Kang Jeong-Mook Park Hyung-Ju Cho |
| author_sort | Dong-Hyeon Kim |
| collection | DOAJ |
| description | Deep learning techniques have been widely applied to classify tree species and segment tree structures. However, most recent studies have focused on the canopy and trunk segmentation, neglecting the branch segmentation. In this study, we proposed a new approach involving the use of the PointNet++ model for segmenting the canopy, trunk, and branches of trees. We introduced a preprocessing method for training LiDAR point cloud data specific to trees and identified an optimal learning environment for the PointNet++ model. We created two learning environments with varying numbers of representative points (between 2048 and 8192) for the PointNet++ model. To validate the performance of our approach, we empirically evaluated the model using LiDAR point cloud data obtained from 435 tree samples scanned by terrestrial LiDAR. These tree samples comprised Korean red pine, Korean pine, and Japanese larch species. When segmenting the canopy, trunk, and branches using the PointNet++ model, we found that resampling 25,000–30,000 points was suitable. The best performance was achieved when the number of representative points was set to 4096. |
| first_indexed | 2024-03-11T02:27:28Z |
| format | Article |
| id | doaj.art-44ecfe096c044766ab5124e7765cd046 |
| institution | Directory Open Access Journal |
| issn | 1999-4907 |
| language | English |
| last_indexed | 2024-03-11T02:27:28Z |
| publishDate | 2023-06-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Forests |
| spelling | doaj.art-44ecfe096c044766ab5124e7765cd0462023-11-18T10:27:11ZengMDPI AGForests1999-49072023-06-01146115910.3390/f14061159Automated Segmentation of Individual Tree Structures Using Deep Learning over LiDAR Point Cloud DataDong-Hyeon Kim0Chi-Ung Ko1Dong-Geun Kim2Jin-Taek Kang3Jeong-Mook Park4Hyung-Ju Cho5Department of Forest Ecology and Protection, Kyungpook National University, Sangju 37224, Republic of KoreaForest ICT Research Center, National Institute of Forest Science, Seoul 02455, Republic of KoreaDepartment of Forest Ecology and Protection, Kyungpook National University, Sangju 37224, Republic of KoreaForest ICT Research Center, National Institute of Forest Science, Seoul 02455, Republic of KoreaForest ICT Research Center, National Institute of Forest Science, Seoul 02455, Republic of KoreaDepartment of Software, Kyungpook National University, Sangju 37224, Republic of KoreaDeep learning techniques have been widely applied to classify tree species and segment tree structures. However, most recent studies have focused on the canopy and trunk segmentation, neglecting the branch segmentation. In this study, we proposed a new approach involving the use of the PointNet++ model for segmenting the canopy, trunk, and branches of trees. We introduced a preprocessing method for training LiDAR point cloud data specific to trees and identified an optimal learning environment for the PointNet++ model. We created two learning environments with varying numbers of representative points (between 2048 and 8192) for the PointNet++ model. To validate the performance of our approach, we empirically evaluated the model using LiDAR point cloud data obtained from 435 tree samples scanned by terrestrial LiDAR. These tree samples comprised Korean red pine, Korean pine, and Japanese larch species. When segmenting the canopy, trunk, and branches using the PointNet++ model, we found that resampling 25,000–30,000 points was suitable. The best performance was achieved when the number of representative points was set to 4096.https://www.mdpi.com/1999-4907/14/6/1159PointNet++segmentationLiDAR point cloud datadeep learning |
| spellingShingle | Dong-Hyeon Kim Chi-Ung Ko Dong-Geun Kim Jin-Taek Kang Jeong-Mook Park Hyung-Ju Cho Automated Segmentation of Individual Tree Structures Using Deep Learning over LiDAR Point Cloud Data Forests PointNet++ segmentation LiDAR point cloud data deep learning |
| title | Automated Segmentation of Individual Tree Structures Using Deep Learning over LiDAR Point Cloud Data |
| title_full | Automated Segmentation of Individual Tree Structures Using Deep Learning over LiDAR Point Cloud Data |
| title_fullStr | Automated Segmentation of Individual Tree Structures Using Deep Learning over LiDAR Point Cloud Data |
| title_full_unstemmed | Automated Segmentation of Individual Tree Structures Using Deep Learning over LiDAR Point Cloud Data |
| title_short | Automated Segmentation of Individual Tree Structures Using Deep Learning over LiDAR Point Cloud Data |
| title_sort | automated segmentation of individual tree structures using deep learning over lidar point cloud data |
| topic | PointNet++ segmentation LiDAR point cloud data deep learning |
| url | https://www.mdpi.com/1999-4907/14/6/1159 |
| work_keys_str_mv | AT donghyeonkim automatedsegmentationofindividualtreestructuresusingdeeplearningoverlidarpointclouddata AT chiungko automatedsegmentationofindividualtreestructuresusingdeeplearningoverlidarpointclouddata AT donggeunkim automatedsegmentationofindividualtreestructuresusingdeeplearningoverlidarpointclouddata AT jintaekkang automatedsegmentationofindividualtreestructuresusingdeeplearningoverlidarpointclouddata AT jeongmookpark automatedsegmentationofindividualtreestructuresusingdeeplearningoverlidarpointclouddata AT hyungjucho automatedsegmentationofindividualtreestructuresusingdeeplearningoverlidarpointclouddata |