基于融合点云数据的马尾松林地单木分割算法研究

doi:10.13466/j.cnki.lczyyj.2024.02.011

摘要/Abstract

摘要：

激光雷达技术在森林资源调查中具有较大优势,但单平台采集的数据往往存在扫描盲区,难以获取完整的森林结构信息。为此,以马尾松林作为研究对象,探究基于融合点云数据的马尾松林单木分割适宜性算法。首先提出一种针对森林样地点云数据融合的方法,然后采用标记控制分水岭算法、距离判别聚类算法和层堆叠算法对马尾松林进行单木分割,并对3种算法的关键参数的选取进行分析,最后提取树高验证融合点云估测森林结构参数的适用性。得出实验结果如下:1)提出的点云融合方法可以有效融合机载和手持激光雷达点云,配准误差为0.054 m;2)3种单木分割算法中,标记控制分水岭算法分割精度最高,总体精度为0.88,高于距离判别聚类算法和层堆叠算法;3)利用标记控制分水岭算法分割的单木提取树高,基于融合点云数据的R²值为0.983 7,RMSE为0.759 6 m,相较于单一点云数据,精度明显提高。研究结果可为多源激光雷达在林业领域的应用以及马尾松林地森林资源管理提供技术支持。

关键词: 单木分割, 融合点云数据, 单木树高, 标记控制分水岭算法, 马尾松

Abstract:

LiDAR technology has a great advantage in forest resources investigation,but the data collected by a single platform often has scanning blind spots,which makes it difficult to obtain complete forest structure information.For this reason,we take the Pinus massoniana forests as the research object and explore the suitability algorithm for individual tree segmentation of Pinus massoniana forests based on fusion point cloud data.We first proposed a method for fusing forest-sample point cloud data.Then,we adopted the marker-controlled watershed algorithm,distance-based clustering algorithm,and the layer stacking algorithm for the Pinus massoniana forests for individual tree segmentation.Finally,tree heights were extracted to verify the applicability of the fusion point cloud for estimating forest structural parameters.The experimental results are as follows:1)The proposed fusion point cloud method can effectively fuse airborne lidar point cloud and hand-held lidar point cloud with a registration error of 0.054 m.2)Among the three tree segmentation algorithms,the marker-controlled watershed algorithm has the highest segmentation accuracy,with an overall accuracy of 0.88,which is higher than the distance-based clustering algorithm and the layer stacking algorithm.3)The extracted tree height of individual tree segmented using the marker-controlled watershed algorithm has an R² of 0.983 7 and an RMSE of 0.759 6 m based on the fusion point cloud data,which is a significant improvement in accuracy compared to single point cloud data.The results of the study can provide technical support for the application of multi-source LiDAR in forestry field and the management of forest resources in Pinus massoniana forests.

Key words: individual tree segmentation, fusion point cloud data, height of individual tree, marker-controlled watershed algorithm, Pinus massoniana

中图分类号:

李炜, 王晓红. 基于融合点云数据的马尾松林地单木分割算法研究[J]. 林草资源研究, 2024,(2): 92-100.

LI Wei, WANG Xiaohong. Individual Tree Segmentation Algorithm of Pinus MassonianaForest Based on Fusion Point Cloud Data[J]. Forest and Grassland Resources Research, 2024,(2): 92-100.

图/表 12

图1

图2

图3

图4

图5

图6

表1

表2

图7

表3

图8

图9

参考文献 23

[1]	SZULECKA J, ZALAZAR E M. Forest plantations in Paraguay:Historical developments and a critical diagnosis in a SWOT-AHP framework[J]. Land Use Policy, 2017,60:384-394.
[2]	陈浩. 马尾松林分质量评价研究:以石阡县与印江县国家储备林为例[D]. 贵阳: 贵州大学, 2022.
[3]	朱泊东, 罗洪斌, 金京, 等. 高郁闭度人工林无人机激光雷达单木分割方法优化[J]. 林业科学, 2022, 58(9):48-59.
[4]	乔俊峰, 周沅桢, 王永, 等. 三维激光扫描测体积技术及其应用进展[J]. 激光与红外, 2021, 51(9):1115-1122.
[5]	王鑫, 潘华志, 罗胜, 等. 机载激光雷达测深技术研究与进展[J]. 海洋测绘, 2019, 39(5):78-82.
[6]	王子斐. 基于手持式移动激光雷达的近自然经营林分结构特征及变化研究[D]. 上海: 华东师范大学, 2021.
[7]	ZHAO Dan, PANG Yong, LI Zengyuan, et al. Isolating individual trees in a closed coniferous forest using small footprint lidar data[J]. International Journal of Remote Sensing, 2014, 35(20):7199-7218.
[8]	惠振阳, 李娜, 程朋根, 等. 基于连通性标记优化的地基LiDAR点云单木分割方法[J]. 中国激光, 2023, 50(6):147-155.
[9]	AYREY E, FRAVER S, KERSHAW J A, et al. Layer stacking:A novel algorithm for individual forest tree segmentation from LiDAR point clouds[J]. Canadian Journal of Remote Sensing, 2017, 43(1):16-27.
[10]	YAN Wanqian, GUAN Haiyan, CAO Lin, et al. A self-adaptive mean shift tree-segmentation method using UAV LiDAR data[J]. Remote Sensing, 2020, 12(3):515.
[11]	ZHAO Xiaoqian, GUO Qinghua, SU Yanjun, et al. Improved progressive TIN densification filtering algorithm for airborne LiDAR data in forested areas[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016,117:79-91.
[12]	袁志聪. 基于Harris特征的点云配准方法研究[D]. 南昌: 东华理工大学, 2019.
[13]	林榆森, 李秋洁, 陈婷. 基于快速点特征直方图的树木点云配准[J]. 中国农机化学报, 2022, 43(2):155-162.
[14]	刘翔宇, 王健, 王效盖, 等. 基于3DSIFT特征点的改进ICP点云配准算法[J]. 应用激光, 2023, 43(11):153-160.
[15]	LIU Li, LIM S, SHEN Xuesong, et al. A hybrid method for segmenting individual trees from airborne lidar data[J]. Computersand Electronics in Agriculture, 2019,163:104871.
[16]	胡迎香, 高红旗, 夏万求, 等. 机载雷达点云亚热带针叶林单木分割探究[J]. 应用激光, 2021, 41(6):1301-1309.
[17]	LIU Kun, SHEN Xin, CAO Lin, et al. Estimating forest structural attributes using UAV-LiDAR data in Ginkgo plantations[J]. ISPRS Jouranalof Photogrammetry and Remote Sensing, 2018,146:465-482.
[18]	PARIS C, KKLBE D, VAN AARDT J, et al. A novel automatic method for the fusion of ALS and TLS LiDAR data for robust assessment of tree crown structure[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(7):3679-3693.
[19]	黄洪宇, 骆钰波, 唐丽玉, 等. 基于单木位置特征的多源树木三维点云配准方法[J]. 林业科学, 2022, 58(11):96-107.
[20]	LIU Qianwei, WANG Jinliang, MA Weifeng, et al. Target-free ULS-TLS point-cloud registration for alpine forest lands[J]. Computers and Electronics in Agriculture, 2021,190:106460.
[21]	苟丙荣. 联合地基LiDAR与无人机LiDAR的白桦蓄积量反演研究[D]. 兰州: 兰州交通大学, 2021.
[22]	段祝庚, 肖化顺, 袁伟湘. 基于离散点云数据的森林冠层高度模型插值方法[J]. 林业科学, 2016, 52(9):86-94.
[23]	魏金龙, 李明阳, 赵邑晨, 等. 基于稀疏型机载激光雷达数据的风景林参数估测[J]. 西北林学院学报, 2021, 36(2):164-171.

插值方式	指标	CHM栅格分辨率
插值方式	指标	0.3 m× 0.3 m	0.4 m× 0.4 m	0.5 m× 0.5 m	0.6 m× 0.6 m
反距离权重	r	1.00	0.97	0.85	0.73
	p	0.47	0.70	0.88	0.87
	F	0.64	0.81	0.86	0.79
克里金	r	0.99	0.96	0.82	0.70
	p	0.45	0.62	0.85	0.85
	F	0.62	0.75	0.83	0.77
不规则三角网	r	0.97	0.96	0.77	0.70
	p	0.45	0.65	0.85	0.83
	F	0.61	0.77	0.81	0.76

距离阈值/m	分割精度
距离阈值/m	r	p	F
1.85	0.85	0.43	0.57
3.97	0.82	0.64	0.72
5.50	0.75	0.66	0.70

层厚度/m	分割精度
层厚度/m	r	p	F
0.2	0.72	0.69	0.70
0.5	0.75	0.73	0.74
1.0	0.71	0.51	0.60