基于移动激光扫描的行道树树冠点云逐点检测

doi:10.12302/j.issn.1000-2006.202202018

国家林草科技领军期刊
中国精品科技期刊
中国高校百佳科技期刊
江苏省新闻出版政府奖期刊奖
RCCSE林学权威期刊（A+）
CSCD核心期刊
Scopus数据库收录期刊
中文核心期刊
SCD核心期刊

作者加群：102861116

微信公众号：南京林业大学学报

高级检索

南京林业大学学报（自然科学版） ›› 2024, Vol. 48 ›› Issue (1): 205-213.doi: 10.12302/j.issn.1000-2006.202202018

基于移动激光扫描的行道树树冠点云逐点检测

李秋洁(), 李相程

南京林业大学机械电子工程学院,江苏南京 210037

收稿日期:2022-02-21 修回日期:2023-06-02 出版日期:2024-01-30 发布日期:2024-01-24
基金资助:
国家自然科学基金项目(31901239)

Pointwise detection of street tree crown point clouds based on mobile laser scanning

LI Qiujie(), LI Xiangcheng

College of Mechanical and Electronic Engineering, Nanjing Forestry University, Nanjing 210037, China

Received:2022-02-21 Revised:2023-06-02 Online:2024-01-30 Published:2024-01-24

摘要/Abstract

摘要：

【目的】针对行道树树冠在线检测问题,研究基于移动激光扫描(mobile laser scanning,MLS)的行道树树冠点云逐点检测方法,构建能够在线、快速、准确检测出行道树树冠点云的高性能树冠检测器,为行道树对靶施药提供基础数据。【方法】应用搭载一个2D激光雷达(light detection and ranging,LiDAR)的MLS系统实时采集街道轮廓线测量数据,从中提取点云三维坐标、一次回波强度和回波次数等属性;构建点云半径为δ的球域搜索方法,实现点云邻域在线快速查询;从待识别点δ球域中提取宽度、深度、高度、维度、密度、次数和强度7类点云局部特征;采用监督学习算法融合点云局部特征、训练树冠检测器,预测待识别点的类别。采集一段长137 m街道的点云数据,开展了邻域搜索方法、监督学习算法、点云局部特征和树冠逐点检测器4个对比实验。【结果】构建的δ球域搜索方法的搜索时间为k-D树法的10.90%;在神经网络(neural network,NN)、支持向量机(support vector machine,SVM)、Boosting和随机森林(random forest,RF)4种监督学习算法中,RF算法得到的树冠检测器分类精度最好;与单类特征相比,组合特征具有更好的泛化性能;本研究方法设计的树冠逐点检测器在检测精度和效率上均优于已有方法,球域半径δ在0.1~1.0 m范围内变化时,测试集F₁分数≥97.74%。【结论】提出的方法能够从实时采集的MLS点云数据中快速、准确地检测出行道树树冠点云,为行道树对靶施药提供数据支撑。

关键词: 对靶施药, 行道树, 树冠点云检测, 逐点分类, 移动激光扫描

Abstract:

【Objective】 To address the problem of online detection of street-tree crowns, this study proposes a pointwise detection method for the point cloud of street-tree crowns based on mobile laser scanning (MLS), and generates a high-performance tree-crown detector capable of online, fast and accurate detection of the point cloud of street-tree crowns to provide basic data for street-tree-targeted spraying. 【Method】 The MLS system equipped with a 2D LiDAR (light detection and ranging) was used to collect the street contour measurement data in real time, and five attributes such as three-dimensional coordinates, primary echo intensity and number of echoes were extracted; a δ spherical neighborhood search method was built to realize online and fast query of point cloud neighborhood; seven kinds of local features of point cloud such as width, depth, height, dimensionality, density, number of echoes and echo intensity were extracted from the spherical domain of the point to be identified; the supervised learning algorithm was used to fuse the local features of point cloud and train a crown detector to predict the category of the point to be identified. 【Result】 The point cloud data of a 137 m long street were collected, and four comparative experiments of neighborhood search methods, supervised learning algorithms, local features of point clouds, and pointwise crown detectors were conducted. The experimental results show that the query time of the proposed δ spherical neighborhood search method is 10.90% of that of k-D tree; among the four supervised learning algorithms of neural network (NN), support vector machine (SVM), Boosting and random forest (RF), the classification performance of the crown detector trained by RF is the best; compared with single kind of features, combined features have better generalization performance; the pointwise crown detector designed by this paper is obviously superior to the existing method in terms of detection accuracy and efficiency. When the radius δ changes within the range of 0.1-1.0 m, the F₁ score on the test set is ≥ 97.74%. 【Conclusion】 The proposed method can quickly and accurately detect street-tree crown point clouds from MLS point-cloud data collected in real time and provide a spray prescription map for targeted spraying applications on street trees.

Key words: targeted spraying, street tree, crown point cloud detection, pointwise classification, mobile laser scanning

中图分类号:

TH112
S718

李秋洁,李相程. 基于移动激光扫描的行道树树冠点云逐点检测[J]. 南京林业大学学报（自然科学版）, 2024, 48(1): 205-213.

LI Qiujie, LI Xiangcheng. Pointwise detection of street tree crown point clouds based on mobile laser scanning[J].Journal of Nanjing Forestry University (Natural Science Edition), 2024, 48(1): 205-213.DOI: 10.12302/j.issn.1000-2006.202202018.

图/表 14

图1

图2

图3

表1

7类点云局部特征"

类别 category	特征 feature
宽度width	σ_x、Δ_x
深度depth	μ_y、σ_y、Δ_y
高度height	μ_z、σ_z、Δ_z
维度dimension	L_λ、P_λ、S_λ、O_λ
密度density	d
次数number	μ_n、σ_n、Δ_n
强度intensity	$μ I 1$ 、 $σ I 1$ 、 $Δ I 1$

表1

图4

表2

图5

图6

表3

表4

表5

表6

图7

表7

参考文献 27

[1]	袁琨, 郭佳, 陆哲明, 等. 篱垣型垂直绿化的减噪能力及其影响因素[J]. 中国城市林业, 2021, 19(4):67-71.
	YUAN K, GUO J, LU Z M, et al. Noise reduction ability of vertical greening of hedgerow and its influencing factors[J]. J Chin Urban For, 2021, 19(4):67-71.DOI: 10.12169/zgcsly.2020.09.10.0001.
[2]	李苹苹, 苗纯萍, 陈玮, 等. 城市街谷行道树对PM_2.5浓度影响的数值模拟研究[J]. 生态学杂志, 2021, 40(12):4044-4052.
	LI P P, MIAO C P, CHEN W, et al. Numerical simulation on the effects of street trees on PM_2.5 concentration in street canyon[J]. Chin J Ecol, 2021, 40(12):4044-4052.DOI: 10.13292/j.1000-4890.202111.036.
[3]	蒋应红. 国内外城市行道树综述[J]. 中国市政工程, 2020(6):4-6,109.
	JIANG Y H. A summary of urban street trees at home & abroad[J]. China Munic Eng, 2020(6):4-6, 109.DOI: 10.3969/j.issn.1004-4655.2020.06.002.
[4]	王伟. 行道树栽植养护技术[J]. 河南农业, 2019(35):33,35.
	WANG W. Planting and maintenance technology of street trees[J]. Agric Henan, 2019(35): 33,35.DOI: 10.15904/j.cnki.hnny.2019.35.017.
[5]	商艳上, 苏田, 李臻, 等. 园林行道树复壮技术[J]. 现代农业科技, 2021(12):176-177.
	SHANG Y S, SU T, LI Z, et al. Rejuvenation technology of garden street trees[J]. Modern Agricultural Science and Technology, 2021(12):176-177.DOI: 10.3969/j.issn.1007-5739.2021.12.071.
[6]	李秋洁, 童岳凯, 薛玉玺, 等. 基于YOLACT的行道树靶标点云分割方法[J]. 林业工程学报, 2022, 7(4):144-150.
	LI Q J, TONG Y K, XUE Y X, et al. Point cloud segment method for street tree target based on YOLACT[J]. J Fore Eng, 2022, 7(4):144-150.DOI:10.13360/j.issn.2096-1359.202110034.
[7]	白秋薇, 张信, 罗红品, 等. 设施果园自动对靶精准变量施肥控制系统[J]. 农业工程学报, 2021, 37(12):28-35.
	BAI Q W, ZHANG X, LUO H P, et al. Control system for auto-targeting precision variable-rate fertilization of fruit trees in a greenhouse orchard[J]. Trans Chin Soc Agric Eng, 2021, 37(12):28-35.DOI: 10.11975/j.issn.1002-6819.2021.12.004http://www.tc.
[8]	郑加强, 徐幼林. 环境友好型农药喷施机械研究进展与展望[J]. 农业机械学报, 2021, 52(3):1-16.
	ZHENG J Q, XU Y L. Development and prospect in environment-friendly pesticide sprayers[J]. Trans Chin Soc Agric Mach, 2021, 52(3):1-16.DOI: 10.6041/j.issn.1000-1298.2021.03.001.
[9]	焦祥, 张慧春, 郑加强, 等. 基于农林植物表型的智能喷雾机械研究进展[J]. 世界林业研究, 2020, 33(5):42-46.
	JIAO X, ZHANG H C, ZHENG J Q, et al. Research progress in intelligent sprayers based on phenotype of agricultural and forestry plants[J]. World For Res, 2020, 33(5):42-46.DOI: 10.13348/j.cnki.sjlyyj.2020.0021.y.
[10]	袁鹏成, 李秋洁, 邓贤, 等. 基于LiDAR的对靶喷雾实时控制系统设计与试验[J]. 农业机械学报, 2020, 51(S1):273-280.
	YUAN P C, LI Q J, DENG X, et al. Design and experiment of real-time control system for target spraying based on LiDAR[J]. Trans Chin Soc Agric Mach, 2020, 51(S1):273-280.
[11]	SULTAN M M, ZAHID A, HE L, et al. Development of a LiDAR-guided section-based tree canopy density measurement system for precision spray applications[J]. Comput Electron Agric, 2021, 182:106053.DOI: 10.1016/j.compag.2021.106053.
[12]	CHEN L M, ZHU H P, HORST L, et al. Management of pest insects and plant diseases in fruit and nursery production with laser-guided variable-rate sprayers[J]. Hort Science, 2021, 56(1):94-100.DOI: 10.21273/hortsci15491-20.
[13]	SAFAIE A H, RASTIVEIS H, SHAMS A, et al. Automated street tree inventory using mobile LiDAR point clouds based on Hough transform and active contours[J]. ISPRS J Photogramm Remote Sens, 2021, 174:19-34.DOI: 10.1016/j.isprsjprs.2021.01.026.
[14]	BIENERT A, GEORGI L, KUNZ M, et al. Automatic extraction and measurement of individual trees from mobile laser scanning point clouds of forests[J]. Ann Bot, 2021, 128(6):787-804.DOI: 10.1093/aob/mcab087.
[15]	XU S, SUN X Y, YUN J Y, et al. A new clustering-based framework to the stem estimation and growth fitting of street trees from mobile laser scanning data[J]. IEEE J Sel Top Appl Earth Obs Remote Sens, 2020, 13:3240-3250.DOI: 10.1109/JSTARS.2020.3001978.
[16]	陆清屿, 李秋洁, 童岳凯, 等. 基于Mask R-CNN的行道树实例分割方法[J]. 林业工程学报, 2021, 6(5):154-160.
	LU Q Y, LI Q J, TONG Y K, et al. Instance segmentation method of street trees based on Mask R-CNN[J]. J Fore Eng, 2021, 6(5):154-160.DOI:10.13360/j.issn.2096-1359.202010011.
[17]	WEINMANN M, WEINMANN M, MALLET C, et al. A classification-segmentation framework for the detection of individual trees in dense MMS point cloud data acquired in urban areas[J]. Remote Sens, 2017, 9(3):277.DOI: 10.3390/rs9030277.
[18]	WEINMANN M, HINZ S, WEINMANN M. A hybrid semantic point cloud classification-segmentation framework based on geometric features and semantic rules[J]. PFG, 2017, 85(3):183-194.DOI: 10.1007/s41064-017-0020-5.
[19]	HACKEL T, WEGNER J D, SCHINDLER K. Fast semantic segmentation of 3D point clouds with strongly varying density[J]. ISPRS Ann Photogramm Remote Sens Spatial Inf Sci, 2016, III-3:177-184.DOI: 10.5194/isprsannals-iii-3-177-2016.
[20]	LANDRIEU L, RAGUET H, VALLET B, et al. A structured regularization framework for spatially smoothing semantic labelings of 3D point clouds[J]. ISPRS J Photogramm Remote Sens, 2017, 132:102-118.DOI: 10.1016/j.isprsjprs.2017.08.010.
[21]	PUENTE I, GONZÁLEZ-JORGE H, ARIAS P, et al. Land-based mobile laser scanning systems:a review[C]// The International Archives of the Photogrammetry Remote Sensing and Spatial Information Sciences, 2011. DOI: 10.5194/isprsarchives-xxxviii-5-w12-163-2011.
[22]	WANG C, WEN C L, DAI Y D, et al. Urban 3D modeling with mobile laser scanning:a review[J]. Virtual Real Intell Hardw, 2020, 2(3):175-212.DOI: 10.1016/j.vrih.2020.05.003.
[23]	李天, 何雄奎, 王志翀, 等. 基于LiDAR技术的喷雾量三维空间分布测试方法[J]. 农业工程学报, 2021, 37(6):42-49.
	LI T, HE X K, WANG Z C, et al. Method for measuring the 3D spatial distribution of spray volume based on LiDAR[J]. Trans Chin Soc Agric Eng, 2021, 37(6):42-49.DOI: 10.11975/j.issn.1002-6819.2021.06.006.
[24]	MADER D, WESTFELD P, MAAS H G. An integrated flexible self-calibration approach for 2D laser scanning range finders applied to the hokuyo UTM-30LX-EW[C]// The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2014.DOI: 10.5194/isprsarchives-xl-5-385-2014.
[25]	WEINMANN M, URBAN S, HINZ S, et al. Distinctive 2D and 3D features for automated large-scale scene analysis in urban areas[J]. Comput Graph, 2015, 49:47-57.DOI: 10.1016/j.cag.2015.01.006.
[26]	DEMANTKÉ J, MALLET C, DAVID N, et al. Dimensionality based scale selection in 3D lidar point clouds[C]// The International Archives of the Photogrammetry Remote Sensing and Spatial Information Sciences, 2011.DOI: 10.5194/isprsarchives-xxxviii-5-w12-97-2011.
[27]	周志华. 机器学习[M]. 北京: 清华大学出版社, 2016.
	ZHOU Z H. Machine learning[M]. Beijing: Tsinghua University Press, 2016.

类别 category	训练集 training set	测试集 test set
道路长度/m street length	33	241
行道树株数 number of street trees	6	44
点云个数 number of points	840 897	6 468 879
树冠点云个数 number of crown points	76 797	886 924
非树冠点云个数 number of non-crown points	764 100	5 581 955

δ/m	搜索时间/s search time
δ/m	本研究方法 proposed method	k-D树 k-D tree
0.1	36.09	408.98
0.2	63.46	689.03
0.3	103.39	1 073.92
0.4	148.79	1 407.73
0.5	202.82	1 924.96
0.6	266.55	2 210.29
0.7	341.11	2 954.06
0.8	428.32	3 475.83
0.9	500.04	4 362.24
1.0	569.89	5 902.45
平均average	266.05	2 440.95

真实类别 actual class	预测类别 predicted class
真实类别 actual class	树冠 crown	非树冠 non-crown
树冠 crown	真正例 (T_P)	假反例 (F_N)
非树冠 non-crown	假正例 (F_P)	真反例 (T_N)

监督学习算法 supervised learning algorithm	算法超参数 algorithm hyperparameter	训练集 training set				测试集 test set
监督学习算法 supervised learning algorithm	算法超参数 algorithm hyperparameter	训练时间/s training time	查准率/% precision	查全率/% recall	F₁分数/% F₁ score	检测时间/s detection time	查准率/% precision	查全率/% recall	F₁分数/% F₁ score
神经网络 NN	hiddenSizes为10	28.69	99.90	99.94	99.92	1.04	98.46	98.62	98.54
支技向量机 SVM	线性核	2.38	87.91	83.52	85.66	0.23	92.05	82.89	87.23
Boosting	NLearn为350	138.61	100.00	100.00	100.00	95.45	99.06	98.89	98.97
随机森林 RF	NumTrees为10	12.56	100.00	100.00	100.00	8.05	99.28	99.28	99.28

特征类别 feature type	特征 feature	训练集training set			测试集test set
特征类别 feature type	特征 feature	查准率/% precision	查全率/% recall	F₁分数/% F₁ score	查准率/% precision	查全率/% recall	F₁分数/% F₁ score
单类特征 single feature	宽度 width	97.70	92.76	95.17	64.01	46.15	53.63
	深度 depth	99.86	99.61	99.73	89.78	72.76	80.38
	高度 height	99.75	98.91	99.33	84.52	77.87	81.06
	维度 dimension	99.17	96.75	97.95	77.88	63.60	70.02
	密度 density	78.01	52.80	62.98	65.51	41.15	50.55
	次数 number	97.37	89.46	93.25	83.79	73.02	78.04
	强度 intensity	99.93	99.85	99.89	78.48	84.69	81.47
组合特征 combined feature	宽度+深度+高度 width+depth+height	100.00	100.00	100.00	99.14	99.00	99.07
	宽度+深度+高度+维度 width+depth+height+dimension	100.00	100.00	100.00	99.14	99.14	99.14
	宽度+深度+高度+维度+密度 width+depth+height+dimension+density	100.00	100.00	100.00	99.11	99.11	99.11
	宽度+深度+高度+维度+密度+次数 width+depth+height+dimension+density+number	100.00	100.00	100.00	99.25	99.25	99.25
	宽度+深度+高度+维度+密度+次数+强度 width+depth+height+dimension+density+number+intensity	100.00	100.00	100.00	99.28	99.28	99.28

基于移动激光扫描的行道树树冠点云逐点检测

Pointwise detection of street tree crown point clouds based on mobile laser scanning

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 27

相关文章 6

编辑推荐

Metrics

本文评价

作者

读者

审稿

方法 method	δ/m	训练集 training set				测试集 test set
方法 method	δ/m	训练时间/s training time	查准率/% precision	查全率/% recall	F₁分数/% F₁ score	平均每帧检测时间/ms average detection time per frame	查准率/% precision	查全率/% recall	F₁分数/% F₁ score
本研究方法 proposed method	0.1	57.71	99.98	99.99	99.98	22.58	98.85	98.20	98.52
	0.2	69.20	100.00	100.00	100.00	28.91	98.98	98.80	98.89
	0.3	88.79	100.00	100.00	100.00	39.56	99.13	98.68	98.90
	0.4	130.62	100.00	100.00	100.00	62.74	99.16	98.72	98.94
	0.5	192.82	100.00	100.00	100.00	96.19	99.23	98.70	98.96
	0.6	247.38	100.00	100.00	100.00	125.53	99.24	98.48	98.86
	0.7	314.85	100.00	100.00	100.00	162.10	99.24	98.00	98.62
	0.8	407.73	100.00	100.00	100.00	211.22	99.22	98.33	98.77
	0.9	466.39	100.00	100.00	100.00	243.49	99.18	97.42	98.29
	1.0	521.49	100.00	100.00	100.00	273.60	99.19	96.33	97.74
k-D树 k-D tree		1 136.03	100.00	100.00	100.00	1 609.68	94.28	88.14	91.11

[1]	贺坤, 王俊洁, 王本耀, 朱海军, 奉树成. 上海市行道树土壤肥力特征及其空间分布[J]. 南京林业大学学报（自然科学版）, 2023, 47(3): 164-172.
[2]	严巍1,3,杨瑞卿1,3,胡永红. 不同栽培基质条件对4种行道树苗木根系生长的影响[J]. 南京林业大学学报（自然科学版）, 2019, 62(04): 192-198.
[3]	熊金鑫,祁慧君,王倩茹,汪倩如,王少华,左文涛,孙圆. 基于i-Tree模型的城市小区行道树生态效益评价[J]. 南京林业大学学报（自然科学版）, 2019, 62(02): 128-136.
[4]	栾以玲,姜志林,阮宏华. 南京市几种行道树年轮重金属元素含量的变化[J]. 南京林业大学学报（自然科学版）, 2009, 52(06): 147-.
[5]	钱能志,薛建辉*,吴永波,任引. 遵义市城区行道树组成结构分析[J]. 南京林业大学学报（自然科学版）, 2005, 48(04): 113-116.
[6]	田如男;薛建辉;潘良;吴菱蓉. 铅胁迫对4种常绿阔叶行道树幼苗细胞膜透性的影响[J]. 南京林业大学学报（自然科学版）, 2004, 47(04): 43-46.