An improved CART model for leaf and wood classification from LiDAR point clouds of Quercus glauca individual trees

doi:10.12302/j.issn.1000-2006.202211006

JOURNAL OF NANJING FORESTRY UNIVERSITY ›› 2024, Vol. 48 ›› Issue (4): 123-131.doi: 10.12302/j.issn.1000-2006.202211006

Special Issue: 专题报道Ⅲ：智慧林业之森林可视化研究

Previous Articles Next Articles

An improved CART model for leaf and wood classification from LiDAR point clouds of Quercus glauca individual trees

PAN Zhengshang¹(), MA Kaisen¹, LONG Yi¹, LAI Zhengui², SUN Hua¹^,^*()

1. Research Center of Forestry Remote Sensing & Information Engineering, Central South University of Forestry and Technology, Key Laboratory of Forestry Remote Sensing Based Big Data & Ecological Security for Hunan Province, Key Laboratory of National Forestry and Grassland Administration on Forest Resources Management and Monitoring in Southern China, Changsha 410004, China
2. Lutou Experimental Forest Farm, Central South University of Forestry and Technology, Yueyang 414000, China

Received:2022-11-04 Revised:2022-12-28 Online:2024-07-30 Published:2024-08-05
Contact: SUN Hua E-mail:20201100030@csuft.edu.cn;sunhua@csuft.edu.cn

Abstract

Abstract:

【Objective】Due to the complex structure and features, traditional classification models for tree branches and leaf point clouds typically face several problems, including poor stability, low accuracy, model overfitting, and high computational costs. In this study, we propose an improved CART (classification and regression tree) model for leaf and branch classification based on Quercus glauca individual tree point cloud data from terrestrial laser LiDAR.【Method】First, the feature descriptor was constructed according to the neighborhood points, and the optimal value of the neighborhood search parameter was then determined. The CART model was improved by gradually introducing variables and adjusting the structure of the decision tree. The classification results of the improved CART model were compared with those of the Logistics regression and K-nearest neighbor (KNN) models.【Result】The accuracy of the improved CART model using the test data increased after introducing the feature descriptors as variables, exceeding that of the Logistics regression and KNN model by 13.1% and 13.6%, respectively. Moreover, the improved CART model exhibited higher accuracy, better stability, and marked reduced model size following the improvement. In particular, the model size was reduced by 99.9% compared with before the improvement, while the data training time was only 51.3% of that before the adjustment. The comprehensive evaluation index of the improved CART model was approximately 0.9 on both trunk and leaf data, with the difference between accuracy on train data and test data lower than 0.001, indicating no overfitting.【Conclusion】The improved CART model has a high accuracy and stability, and achieves good classification results on small samples. This study provides a methodological reference for the accurate and rapid classification of trunk and leaf point clouds from terrestrial laser LiDAR.

Key words: terrestrial laser LiDAR, point cloud classification, point cloud features, classification and regression tree

CLC Number:

S758
TP391

PAN Zhengshang, MA Kaisen, LONG Yi, LAI Zhengui, SUN Hua. An improved CART model for leaf and wood classification from LiDAR point clouds of Quercus glauca individual trees[J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2024, 48(4): 123-131.

Figures/Tables 10

Table 1

Table 2

Computing formulas of feature descriptor"

特征 feature	计算公式 calculation formula	特征 feature	计算公式 calculation formula
线性指数 linearity index	L_λ= $λ 1 - λ 2 λ 1$	各向异性 nisotropy	A_λ= $λ 1 - λ 3 λ 2$
平面性指数 planarity index	P_λ= $λ 2 - λ 3 λ 1$	特征熵 information entropy	E_λ=- $∑ i = 1 3$ λ_i(lnλ_i)
球性指数 sphericity index	S_λ= $λ 3 λ 1$	迹 trace	T_λ= $2 π$ arctan( $∑ i = 1 3$ λ_i)
各向同性 isotropy	O_λ= $λ 1 λ 2 λ 3 3$	曲率 curvature	C_λ= $λ 3 λ 1 + λ 2 + λ 3$

Table 2

Fig. 1

Table 3

Fig. 2

Fig. 3

Table 4

Fig. 4

Table 5

Fig. 5

References 30

[1]	TAO S L, GUO Q H, XU S W, et al. A geometric method for wood-leaf separation using terrestrial and simulated lidar data[J]. Photogram Engng Rem Sens, 2015, 81(10):767-776.DOI: 10.14358/pers.81.10.767.
[2]	杨必胜, 董震. 点云智能研究进展与趋势[J]. 测绘学报, 2019, 48(12):1575-1585.
	YANG B S, DONG Z. Progress and perspective of point cloud intelligence[J]. Acta Geod Cartogr Sin, 2019, 48(12):1575-1585.DOI: 10.11947/j.AGCS.2019.20190465.
[3]	孙圆, 林秀云, 熊金鑫, 等. 基于地面激光强度校正数据的单木枝叶分离[J]. 中国激光, 2021, 48(1):0104001.
	SUN Y, LIN X Y, XIONG J X, et al. Separation of single wood branches and leaves based on corrected TLS intensity data[J]. Chin J Lasers, 2021, 48(1):0104001.DOI: 10.3788/CJL202148.0104001.
[4]	CALDERS K, DISNEY M I, ARMSTON J, et al. Evaluation of the range accuracy and the radiometric calibration of multiple terrestrial laser scanning instruments for data interoperability[J]. IEEE Trans Geosci Remote Sens, 2017, 55(5):2716-2724.DOI: 10.1109/TGRS.2017.2652721.
[5]	夏国芳, 胡春梅, 曹毕铮, 等. 激光入射角度对点云反射强度的影响研究[J]. 激光杂志, 2016, 37(4):11-13.
	XIA G F, HU C M, CAO B Z, et al. Study on the influence of laser incident angle on the reflection intensity of the point cloud[J]. Laser J, 2016, 37(4):11-13.DOI: 10.14016/j.cnki.jgzz.2016.04.011.
[6]	曹伟, 陈动, 史玉峰, 等. 激光雷达点云树木建模研究进展与展望[J]. 武汉大学学报(信息科学版), 2021, 46(2):203-220.
	CAO W, CHEN D, SHI Y F, et al. Progress and prospect of LiDAR point clouds to 3D tree models[J]. Geomat Inf Sci Wuhan Univ, 2021, 46(2):203-220.DOI: 10.13203/j.whugis20190275.
[7]	黄亮, 许文雅, 谭帅. 树木三维点云的枝叶分割方法[J]. 北京测绘, 2022, 36(1):18-22.
	HUANG L, XU W Y, TAN S. 3D point cloud segmentation method for branches and leaves of trees[J]. Beijing Surv Mapp, 2022, 36(1):18-22.DOI: 10.19580/j.cnki.1007-3000.2022.01.004.
[8]	FARMAKIS I, BONNEAU D, HUTCHINSON D J, et al. Supervoxel-based multi-scale point cloud segmentation using fnea for object-oriented rock slope classification using TLS[J]. Int Arch Photogramm Remote Sens Spatial Inf Sci, 2020, 43:1049-1056.DOI: 10.5194/isprs-archives-xliii-b2-2020-1049-2020.
[9]	林筱涵, 李爱农, 边金虎, 等. 基于网络图的地基激光雷达复杂树木点云枝叶分离方法[J]. 遥感技术与应用, 2022, 37(1):161-172.
	LIN X H, LI A N, BIAN J H, et al. A method for separating leaf and wood components of complex tree point cloud data based on network graph with terrestrial laser scanning[J]. Remote Sens Technol Appl, 2022, 37(1):161-172.DOI: 10.11873/j.issn.1004-0323.2022.1.0161.
[10]	WANG D, MOMO T S, CASELLA E. LeWoS:a universal leaf-wood classification method to facilitate the 3D modelling of large tropical trees using terrestrial LiDAR[J]. Me-thods Ecol Evol, 2020, 11(3):376-389.DOI: 10.1111/2041-210x.13342.
[11]	WANG D, HOLLAUS M, PFEIFER N. Feasibility of machine learning methods for separating wood and leaf points from terrestrial laser scanning data[J]. ISPRS Ann Photogramm Remote Sens Spatial Inf Sci, 2017, IV-2/W4:157-164.DOI: 10.5194/isprs-annals-iv-2-w4-157-2017.
[12]	VICARI M B, DISNEY M, WILKES P, et al. Leaf and wood classification framework for terrestrial LiDAR point clouds[J]. Methods Ecol Evol, 2019, 10(5):680-694.DOI: 10.1111/2041-210x.13144.
[13]	QI C R, YI L, SU H, et al. PointNet++:deep hierarchical feature learning on point sets in a metric space[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. December,2017, Long Beach,California,USA: ACM, 2017:5105-5114.DOI: 10.5555/3295222.3295263.
[14]	赵江洪, 窦新铜, 曹月娥, 等. 一种基于分割结果实现三维点云分类的方法[J]. 测绘科学, 2022, 47(3):85-95.
	ZHAO J H, DOU X T, CAO Y E, et al. A method for 3D point cloud classification based on segmentation results[J]. Sci Surv Mapp, 2022, 47(3):85-95.DOI: 10.16251/j.cnki.1009-2307.2022.03.012.
[15]	ZHOU J J, WEI H Q, ZHOU G Y, et al. Separating leaf and wood points in terrestrial laser scanning data using multiple optimal scales[J]. Sensors, 2019, 19(8):1852.DOI: 10.3390/s19081852.
[16]	汪献义, 邢艳秋, 尤号田, 等. 基于近邻几何特征的TLS林分点云分类研究[J]. 北京林业大学学报, 2019, 41(6):138-146.
	WANG X Y, XING Y Q, YOU H T, et al. TLS point cloud classification of forest based on nearby geometric features[J]. J Beijing For Univ, 2019, 41(6):138-146.DOI: 10.13332/j.1000-1522.20180308.
[17]	麻卫峰, 王金亮, 张建鹏, 等. 一种改进法向量估算的点云特征提取[J]. 测绘科学, 2021, 46(11):84-90,146.
	MA W F, WANG J L, ZHANG J P, et al. Feature extraction from point cloud based on improved normal vector[J]. Sci Surv Mapp, 2021, 46(11):84-90,146.DOI: 10.16251/j.cnki.1009-2307.2021.11.012.
[18]	梁冲. 面向树木特征提取的室外场景点云数据语义分类[D]. 南京: 南京师范大学, 2021.
	LIANG C. Semantic classification of outdoor scenic spot pointcloud data for tree feature extraction[D]. Nanjing: Nanjing Normal University, 2021.
[19]	ZHU X, SKIDMORE A K, DARVISHZADEH R, et al. Foliar and woody materials discriminated using terrestrial LiDAR in a mixed natural forest[J]. Int J Appl Earth Obs Geoinf, 2018, 64:43-50.DOI: 10.1016/j.jag.2017.09.004.
[20]	卢晓艺, 云挺, 薛联凤, 等. 基于树木激光点云的有效特征抽取与识别方法[J]. 中国激光, 2019, 46(5):411-422.
	LU X Y, YUN T, XUE L F, et al. Effective feature extraction and identification method based on tree laser point cloud[J]. Chin J Lasers, 2019, 46(5):411-422.DOI: 10.3788/CJL201946.0510002.
[21]	PATEL H H, PRAJAPATI P. Study and analysis of decision tree based classification algorithms[J]. Ijcse, 2018, 6(10):74-78.DOI: 10.26438/ijcse/v6i10.7478.
[22]	王艳秋, 徐传飞, 于戈, 等. 一种面向不确定对象的可见k近邻查询算法[J]. 计算机学报, 2010, 33(10):1943-1952.
	WANG Y Q, XU C F, YU G, et al. Visible k-nearest neighbor queries over uncertain data[J]. Chin J Comput, 2010, 33(10):1943-1952.DOI: 10.3724/SP.J.1016.2010.01943.
[23]	刘刚, 尹一涵, 郑智源, 等. 基于三维点云的群体樱桃树冠层去噪和配准方法[J]. 农业机械学报, 2022, 53(S2):1-11.
	LIU G, YIN Y H, ZHENG Z Y, et al. Denoising and registration method of group cherry trees canopy based on 3D point cloud[J]. Trans Chin Soc Agric Mach, 2022, 53(S2):1-11.DOI: 10.6041/j.issn.1000-1298.2022.s2.021.
[24]	张卉冉, 董震, 杨必胜, 等. 点云压缩研究进展与趋势[J]. 武汉大学学报(信息科学版), 2023, 48(2):192-205.
	ZHANG H R, DONG Z, YANG B S, et al. Progress and perspectives of point cloud compression[J]. Geomatics and Information Science of Wuhan University, 2023, 48(2):192-205.DOI: 10.13203/j.whugis20210103.
[25]	CHENG X, ZENG M, LIN J P, et al. Efficient L0 resampling of point sets[J]. Comput Aided Geom Des, 2019, 75:101790.DOI: 10.1016/j.cagd.2019.101790.
[26]	DONG Z, YANG B S, LIANG F X, et al. Hierarchical registration of unordered TLS point clouds based on binary shape context descriptor[J]. ISPRS J Photogramm Remote Sens, 2018, 144:61-79.DOI: 10.1016/j.isprsjprs.2018.06.018.
[27]	张继超, 刘宁, 宋伟东, 等. 一种特征选择的全极化雷达影像分类方法[J]. 测绘科学, 2022, 47(6):127-134.
	ZHANG J C, LIU N, SONG W D, et al. A fully polarimetric radar image classification method based on feature selection[J]. Sci Surv Mapp, 2022, 47(6):127-134.DOI: 10.16251/j.cnki.1009-2307.2022.06.016.
[28]	卢华清, 伍吉仓, 张子健. 面向地基激光点云的树木枝叶分离[J]. 中国激光, 2022, 49(23):156-167.
	LU H Q, WU J C, ZHANG Z J. Tree branch and leaf separation facing TLS point cloud[J]. Chin J Lasers, 2022, 49(23):156-167. DOI: 10.3788/CJL.202249.2310001.
[29]	BECKER C, ROSINSKAYA E, HÄNI N, et al. Classification of aerial photogrammetric 3D point clouds[J]. Photogramm Eng Remote Sensing, 2018, 84(5):287-295.DOI: 10.14358/pers.84.5.287.
[30]	邢艳秋, 蔡硕, 汪献义. 采用多尺度近邻体素特征的TLS林分点云分类[J]. 遥感信息, 2021, 36(5):1-7.
	XING Y Q, CAI S, WANG X Y. TLS point cloud classification of forest by multi-scale neighbor voxel features[J]. Remote Sens Inf, 2021, 36(5):1-7.DOI: 10.3969/j.issn.1000-3177.2021.05.001.

Related Articles 15

[1]	YI Jing, MA Kaisen, XIANG Jianping, TANG Jie, JIANG Fugen, CHEN Song, SUN Hua. Research on TLS single tree detection method based on point cloud slicing combined with clustering algorithm [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2024, 48(4): 113-122.
[2]	YAN Yu, LI Qiujie. Online segmentation method of target point cloud in roadside tree [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2024, 48(4): 141-149.
[3]	ZHAO Yugang, LIU Wenping, ZHOU Yan, CHEN Riqiang, ZONG Shixiang, LUO Youqing. UAV forestry land-cover image segmentation method based on attention mechanism and improved DeepLabV3+ [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2024, 48(4): 93-103.
[4]	NIU Hongjian, LIU Wenping, CHEN Riqiang, ZONG Shixiang, LUO Youqing. Dehaze algorithm for woodland UAV images based on Resnet [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2024, 48(2): 175-18.
[5]	GAO Xieyu, DONG Lihu, HAO Yuanshuo. Effects of thinning on Larix olgensis plantation stem form based on TLS [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2023, 47(6): 85-94.
[6]	LI Shuangxian, LU Xin, Duojie Cairen, ZHANG Huaiqing, XUE Lianfeng, YUN Ting. A novel approach for leaf area retrieval from terrestrial laser scanned points [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2023, 47(5): 28-38.
[7]	ZHANG Huacong, TAN Xinjian, YU Longhua, LI Yueqiao, CHEN Yongfu, LIU Ren, ZHANG Huaiqing. Crown segmentation of CHM based on the enhanced frost local filtering and distance map reconstruction [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2023, 47(5): 9-18.
[8]	SUN Yu, LI Fengri, XIE Longfei, DONG Lihu. Construction of the stand-level biomass model of Larix olgensis plantations based on stand and topographic factors [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2023, 47(3): 129-136.
[9]	TANG Yiren, JIA Weiwei, WANG Fan, SUN Yuman, ZHANG Ying. Constructing a biomass model of Larix olgensis primary branches based on TLS [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2023, 47(2): 130-140.
[10]	OU Jiande. Effects of forking on the crown morphology, growth and form quality of Taxus chinensis var. mairei [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2023, 47(2): 87-94.
[11]	WANG Fan, JIA Weiwei, TANG Yiren, LI Dandan. Extraction of crown radius and development of crown profile model of Pinus koraiensis based on TLS [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2023, 47(1): 13-22.
[12]	CHENG Xiaofei, WU Gang. Locating individual tree from high resolution satellite images based on CV model [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2022, 46(5): 143-151.
[13]	NIE Luyi, DONG Lihu, LI Fengri, MIAO Zheng, XIE Longfei. Construction of taper equation for Larix olgensis based on two-level nonlinear mixed effects model [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2022, 46(3): 194-202.
[14]	ZHAO Qing, HUANG Fei, CHEN Xiaohui, LIN Yuying, QIU Rongzu, WU Zhilong, HU Xisheng. Spatiotemporal variations and a driving force analysis of arbor forest loss [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2022, 46(2): 227-235.
[15]	XIN Shidong, JIANG Lichun, MU Lin. Predictive model of stand tree layer additive carbon storage of Korean pine plantation in Heilongjiang Province, China [J]. JOURNAL OF NANJING FORESTRY UNIVERSITY, 2022, 46(1): 115-121.

Metrics

Comments

www.nldxb.njfu.edu.cn

Edited ＆ Published by Editorial Department of Nanjing Forestry University(Natural Sciences Edition)
Address： No.159 Longpan Road,Nanjing 210037 Jiangsu,P.R.China
E-mail ：xuebao@njfu.edu.cn , xuebao@njfu.com.cn
Telephone +86-25-85428247,85427076
Distributed by China International Book Trading Corporation P.O. Box 399, Beijing ,P.R. China

类型 data type	株数 number of trees	树叶点 points of leaf	树干点 points of trunk
训练数据 train data	40	17 072 431	15 749 444
测试数据 test data	10	4 089 535	5 619 785
合计 total	50	21 161 966	21 369 229

实际类别 true type	树叶 leaf	树干 trunk
正确分类 classified correctly	树叶正估点 T_l	树干正估点 T_t
错误分类 classified incorrectly	树叶错估点 F_l	树干错估点 F_t

分类模型 classification model	特征变量数 number of features	最大深度 max depth	节点数 number of node	训练时间/s training time	模型大小/kB model size	准确率 precision
CART	3	111	12 005 413	535.9	864 391	0.722
改进CART improving CART	5	75	4 707 427	704.3	338 936	0.824
调整改进CART adjusting and improving CART	5	12	7 591	361.3	548	0.901

模型 model	数据 data	数据类型 data type	正确分类点 classified correctly points	错误分类点 classified incorrectly points	准确率 precision	精确度 accuracy	召回率 recall	综合评价指标 F-score
LR	训练数据	树干	14 020 775	3 467 492	0.801	0.821	0.801	0.811
		树叶	12 281 952	3 051 656		0.780	0.801	0.790
	测试数据	树干	3 514 156	1 405 210	0.796	0.859	0.714	0.780
		树叶	4 214 575	575 379		0.750	0.880	0.809
KNN	训练数据	树干	14 579 330	2 011 562	0.863	0.854	0.879	0.866
		树叶	13 737 882	2 493 101		0.872	0.846	0.859
	测试数据	树干	3 358 470	1 292 731	0.792	0.821	0.722	0.768
		树叶	4 327 054	731 065		0.770	0.855	0.810
CART	训练数据	树干	17 071 982	724	1	1	1	1
		树叶	15 748 720	449		1	1	1
	测试数据	树干	3 165 326	1735 346	0.726	0.774	0.646	0.704
		树叶	3 884 439	924 209		0.691	0.808	0.745
改进 CART improving CART	训练数据	树干	17 071 993	724	1	1	1	1
		树叶	15 748 720	438		1	1	1
	测试数据	树干	3 491 232	1 056 703	0.830	0.854	0.768	0.808
		树叶	4 563 082	598 303		0.812	0.884	0.846
调整改进 CART adjusting and improving CART	训练数据	树干	14 506 710	776 175	0.898	0.850	0.949	0.897
		树叶	14 973 269	2 565 721		0.951	0.854	0.900
	测试数据	树干	3 476 917	351 018	0.900	0.850	0.908	0.878
		树叶	5 268 767	612 618		0.938	0.895	0.916

An improved CART model for leaf and wood classification from LiDAR point clouds of Quercus glauca individual trees

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 30

Related Articles 15

Recommended Articles

Metrics

Comments

Author

Reader

Reviewer