Extraction of the Measuring Parameters of Pinus armandii Man-Made Forest Based on Visible Image of UAV

doi:10.13466/j.cnki.lyzygl.2021.02.015

Abstract

Abstract:

Thanks to the portability of unmanned aerial vehicles (UAV)in monitoring of forest resources,this paper uses RGB tri-band image of UAV to extract the forest measuring parameters (number of stems,height of tree and stand volume)and perform precision validation.Taking P.armandii man-made forest as research object,RGB image of UAV as main information source and based on five individual-tree positioning of 0.08hm² P.armandii man-made forest sample plots in the early stage,this paper adopts canopy height model (CHM)maximum method and point cloud segmentation method to extract the measuring parameters of P.armandii and establishes a binary volume model of individual-tree in P.armandii man-made forest via RGB image of UAV.The results show that:(1)The results of individual-tree number and tree-height segmentation indicate that CHM maximum segmentation method has a higher precision than point cloud segmentation method,and the segmentation accuracy of the number of stems is 87.17% and 80.79%,respectively;the segmentation accuracy of the height of tree is 0.71 and 0.69,respectively.(2)The determination coefficient (R²)of the binary volume model established according to individual-tree crown breadth and tree-height extracted via CHM maximum method is 0.94,and its root-mean-square error (RMSE)is 0.033 8m³.The results are compared with those of the measured stock volume in standard land based on P.armandii of Yunnan Province binary volume table,the monitoring accuracy of stand volume of the five sample plots based on RGB data of UAV is 79.72%,81.64%,83.57%,82.49% and 80.28%,respectively,with the average accuracy reaching 81.54%.During the extraction of the measuring parameters of P.armandii man-made forest via RGB image of UAV,CHM maximum method is superior to point cloud segmentation method,and the binary volume model of tree height and crown breadth established by it can be used for UAV remote sensing in monitoring P.armandii monolayer man-made forest.

Key words: RGB image of UAV, CHM segmentation, point cloud segmentation, forest measuring parameter, P.armandii man-made forest

CLC Number:

YUAN Zijian, SHU Qingtai, ZHAO Hongying, WANG Keren, TAN Dehong. Extraction of the Measuring Parameters of Pinus armandii Man-Made Forest Based on Visible Image of UAV[J]. FOREST RESOURCES WANAGEMENT, 2021, 0(2): 110-116.

Figures/Tables 10

Fig.1

Tab.1

Fig.2

Fig.3

Tab.2

Estimation equation of binary volume model

模型序号	模型
1	V= $b 0$ + $b 1 D 2$ H
2	V= $b 0 D 2$ + $b 1 D 2$ H
3	V= $b 0 D b 1 H b 2$
4	V= $b 0$ + $b 1 D 2$ + $b 2 D 2$ H
5	V= $b 0$ + $b 1 H$ + $b 2 D 2$ H
6	V= $b 0$ + $b 1 H$ + $b 2 D 2$

Tab.2

Tab.3

Fig.4

Fig.5

Tab.4

Fitting results of binary volume model of P.armandi in State-owned forest farm of Shilin county

模型	参数估计值			R²	RMSE
模型	$b 0$	$b 1$	$b 2$	R²	RMSE
1	0.010232	0.000029		0.811985	0.0457
2	0.000132	0.000054		0.752549	0.0513
3	0.000073	2.001569	0.788883	0.937843	0.0338
4	-0.097267	0.000045	0.000015	0.773348	0.0475
5	0.987542	-0.000736	0.000035	0.717401	0.0568
6	-0.097651	0.000879	-0.000508	0.883084	0.0397

Tab.4

Tab.5

References 20

[1]	李宇昊. 无人机在林业调查中的应用实验[J]. 林业资源管理, 2007(4):69-73.
[2]	李兆琨. 林业调查规划设计在林业生产建设中的作用分析[J]. 绿色科技, 2017(5):99-100.
[3]	林金洪, 潘峰. 林业调查规划设计与林业生产建设的关系探讨[J]. 南方农业, 2016(12):120-121.
[4]	白芝兵, 张洪江, 任改. 重庆四面山森林植物群落数量分类和排序[J]. 安徽农业科学, 2012(30):14839-14843.
[5]	黎燕琼, 郑绍伟, 陈俊华, 等. 成都市中心城区城市森林树木的数量结构特征[J]. 森林与环境学报, 2018(1):71-75
[6]	李旺, 牛铮, 王成, 等. 机载LIDAR数据估算样地和单木尺度森林地上生物量[J]. 遥感学报, 2015(4):669-679.
[7]	刘莉. 森林植物种质资源价值评估的研宄[D]. 哈尔滨: 东北林业大学, 2013.
[8]	王参福, 余于文, 汪林清, 等. 论森林资源数量与质量的辩证关系——以江西省森林资源为例[J]. 江西林业科技, 2001(2):62-65.
[9]	郑治刚, 李留瑜, 付秀山. 论森林的数量分布结构[J]. 系统工程理论与实践, 1996(4):82-89.
[10]	滕建生. 林业调查规划设计在林业生产建设中的地位和作用[J]. 农业与技术, 2015(4):77-78.
[11]	唐景华, 陈育娟. 林业调查规划设计在林业生产建设中的重要性分析[J]. 现代园艺, 2015(20):227-228.
[12]	张艳丽. 杭州市典型城市森林类型生态保健功能研究[D]. 北京: 中国林业科学研宄院, 2013.
[13]	罗仙仙, 亢新刚. 森林资源综合监测研究综述[J]. 浙江林学院学报, 2008(6):803-809.
[14]	陈述彭. “数字地球”与对地观测[J]. 地理学报, 2000,55(1):9-13. doi: 10.11821/xb200001001
[15]	宫鹏. 遥感生态测量学进展[J]. 自然资源学报, 1999(10):313-317.
[16]	王二丽, 李凤日, 王敏, 等. 基于大比例尺航片的针叶树种冠幅的提取[J]. 森林工程, 2011,27(6):16-22.
[17]	Popescu S C, Wynee R H, Scrivani J A. Fusion of small footprint LiDAR and multispectral data to estimate plot-level volume and biomass in deciduous and pine forests in Virginia,USA[J]. Forest Science, 2004,50(4):551-565.
[18]	Koukoulas S, Blackbum G A. Mapping individual tree location,height and species in broadleaved deciduous forest using airbome LiDAR and multi spectral remotely sensed data[J]. International Journal of Romote Sensing. 2005,26(3):431-455.
[19]	Kwak Doo-Ahn, Lee Woo-kyun, Lee Jun-Hak, et al. Detection of individual trees and estimation of tree height using LiDAR data[J]. Journal of Forest Research., 2007(12):425-434.
[20]	宋永俊. 云南省林业数表编制现状分析与对策建议[J]. 林业资源管理, 2015(6):37-40.

样地数/块	胸径/cm				树高/m				冠幅/m
样地数/块	林分平均胸径	最大值	最小值		林分平均高	最大值	最小值		林分平均冠幅	最大值	最小值
5	19.8	10.6	38.2		17.71	23.27	9.24		3.8	7.8	1.1

单木测量方法	未分离株数	过度分离株数	正确分离株数	总株数
基于CHM提取单木	25	28	360	413
基于点云提取单木	34	44	328	406
地面实测单木	0	0	443	443

样地	实测蓄积量/m³	基于CHM提取蓄积量/m³	估测精度/%	基于点云提取蓄积量/m³	估测精度/%
1	11.5553	9.2129	79.72	8.4331	73.12
2	15.7498	12.8581	81.64	11.1004	70.48
3	18.3729	15.3452	83.57	13.2137	71.92
4	18.8449	15.5451	82.49	13.8396	73.44
5	24.3454	19.5444	80.28	17.4386	71.63