Chinese Journal of Lasers, Volume. 51, Issue 8, 0810001(2024)
Point Clouds Classification and Segmentation for Nursery Trees Based on Improved PointNet++ Model
Figures & Tables(16)
Fig. 2. Acquisition of relative coordinates of neighboring points for each local area
Fig. 6. Nursery for data collection. (a) Whole scene of nursery; (b) part of scene of nursery
Fig. 7. Livox Horizon laser sensor and acquired point clouds. (a) Livox Horizon laser sensor; (b) acquired point clouds
Fig. 8. Seven common kinds of landscape trees. (a) Osmanthus fragrans; (b) Malus halliana; (c) cherry plum; (d) Acer palmatum; (e) Chimonanthus praecox; (f) loquat tree; (g) Chinese holly
Fig. 9. Schematic diagram of segmented point clouds of seven tree species. (a) Osmanthus fragrans; (b) Malus halliana; (c) cherry plum; (d) Acer palmatum; (e) Chimonanthus praecox; (f) loquat tree; (g) Chinese holly
Fig. 11. Examples of visualization of segmentation results using PointNet, PointNet++, and proposed model (white boxes denote wrong predicted points). (a) Ground truth; (b) segmentation results of PointNet; (c) segmentation results of PointNet++; (d) segmentation results of proposed model
Table 1. Laser parameters
View tableTable 1. Laser parameters
Parameter Value Detection range /m 260 Range error /cm 2 Angle error /（°） 0.05 Field-of-view （FOV） /（°） 81.7×25.1 Data rate /（point·s-1） 240000
Table 2. Numbers of samples, point cloud groups, and groups in training and testing sets for different types of trees
View tableTable 2. Numbers of samples, point cloud groups, and groups in training and testing sets for different types of trees
Category Number of trees Number of point cloud groups Number of point cloud groups in training set Number of point cloud groups in testing set Osmanthus fragrans 120 179 126 53 Malus halliana 144 283 201 82 Cherry plum 164 287 201 86 Acer palmatum 151 290 203 87 Chimonanthus praecox 158 290 203 87 Loquat tree 146 215 151 64 Chinese holly 163 284 199 85
Table 3. Software and hardware parameters for experiment
View tableTable 3. Software and hardware parameters for experiment
Parameter Content CPU Intel（R） Xeon（R） Gold 6226R GPU NVIDIA RTX3090 Cuda Cuda11.7 Data processing tools Python3.8， PyCharm2020 Deep learning framework Pytorch
Table 4. Testing results of classification on self-made dataset
View tableTable 4. Testing results of classification on self-made dataset
Model OA mAcc PointNet 81.25 80.00 PointNet++ 89.71 89.44 Ours （SE） 90.33 92.39 Ours （CBAM） 90.83 92.28 Ours （CA） 91.33 93.13 Ours 92.50 94.22
Table 5. Testing results of segmentation on self-made dataset
View tableTable 5. Testing results of segmentation on self-made dataset
Model mIoU Precision Recall F1 score PointNet 76.95 70.71 76.01 73.01 PointNet++ 83.85 85.35 92.40 88.47 Ours （SE） 87.08 89.32 94.86 91.45 Ours （CBAM） 87.61 88.56 94.29 91.17 Ours （CA） 89.89 88.29 94.51 91.12 Ours 89.09 90.09 95.44 92.59
Tools
Get Citation
Copy Citation Text
Jie Xu, Hui Liu, Yue Shen, Guanxue Yang, Hao Zhou, Siyuan Wang. Point Clouds Classification and Segmentation for Nursery Trees Based on Improved PointNet++ Model[J]. Chinese Journal of Lasers, 2024, 51(8): 0810001
Paper Information
Category: remote sensing and sensor
Received: Jul. 4, 2023
Accepted: Sep. 5, 2023
Published Online: Mar. 29, 2024
The Author Email: Liu Hui (amity@ujs.edu.cn)
CSTR:32183.14.CJL230989
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20