Tree Level Monitoring of Pine Wilt Disease Based on UAV Multispectral Imagery

doi:10.13466/j.cnki.lyzygl.2022.05.014

Abstract

Abstract:

Pine wilt disease(PWD)is one of the most harmful forest diseases,and there is an urgent need to adopt accurate monitoring means to determine the number and location of diseased trees for efficient prevention and control of PWD.In this study,the image of PWD epidemic area in Zhongcheng Town,Rongjiang County,Guizhou Province was obtained by using multi spectral UAV,and the multi spectral UAV and its derived point cloud were used as data sources.Firstly,localization identification and crown profile segmentation of individual trees in the study area were performed by point cloud segmentation algorithm.The spectral features were then extracted in segmentation units and the best feature set was filtered by a combination of random forest and recursive feature elimination(RF-RFE).Finally,random Forest(RF)and support vector machine(SVM)detection models were constructed based on screening feature sets,and the model detection performance was evaluated. At the same time,the RF and the SVM were used to invert the disease susceptibility in the study area and draw the spatial distribution map of PWD.The following key results were obtained:1)The individual trees segmentation based on photogrammetric point clouds was effective,with an overall F-score value of 82.21%.The OA and Kappa of the RF model constructed after feature screening were 84.4% and 0.74,respectively,and the SVM was 76.09% and 0.66.2)The F-score for RF were 78.43%,69.23%,83.33% and 94.12%,SVM were 80.7%,55.81%,70.18% and 84.13% for the four stages of tree health,early,middle,and late detection,respectively.The comprehensive comparison of the detection performance of RF was the best.The study pointed out that it was feasible to use the combination of UAV multispectral image and photographic measurement point cloud for individual trees scale monitoring of PWD.The study aimed to provide a reference for low-cost and accurate remote sensing monitoring of PWD.

Key words: pine wilt disease, UAV multispectral imagery, photogrammetric point cloud, individual tree crown delineation

CLC Number:

WANG Bu, TAN Wei, WANG Guilin, PU Xiuqing. Tree Level Monitoring of Pine Wilt Disease Based on UAV Multispectral Imagery[J]. FOREST RESOURCES WANAGEMENT, 2022, (5): 107-117.

Figures/Tables 10

Fig.1

Tab.1

Vegetation indices selected in this study

植被指数	公式	参考文献
大气阻抗植被指数(ARVI)	$(N I R - 2 R + B) / (N I R + 2 R + B)$	[23]
大气阻抗植被指数(ARVI2)	$0.17 (N I R - R) / (N I R + R) - 0.18$	[23]
叶绿素植被指数(CVI)	$(N I R × R) / G 2$	[24]
差值植被指数(DVI)	$N I R - R$	[25]
增强植被指数(EVI)	$2.5 (N I R - R) / (N I R + 6 R - 7.5 B + 1)$	[26]
增强植被指数(EVI2)	$2.4 (N I R - R) / (N I R + R + 1)$	[27]
增强植被指数(EVI2-2)	$2.5 (N I R - R) / (N I R + 2.4 R + 1)$	[28]
绿色抗大气植被指数(GARI)	$[N I R - (G - B + R)] / [N I R - (G + B - R)]$	[23]
蓝绿色归一化植被指数(GBNDVI)	$[N I R - G + B] / [N I R + G + B]$	[29]
近红-绿差值植被指数(GDVI)	$N I R - G$	[30]
绿色归一化植被指数(GNDVI)	$(N I R - G) / (N I R + G)$	[29]
绿红色归一化植被指数(GRNDVI)	$[N I R - (G + R)] / [N I R + (G + R)]$	[29]
改进型土壤调整植被指数(MSAVI)	$[2 N I R + 1 - 2 N I R + 1 2 - 8] / 2$	[31]
归一化植被指数(NDVI)	$(N I R - R) / (N I R + R)$	[30]
红边归一化植被指数(NDRE)	$(R e d E d g e - R) / (R e d E d g e + R)$	[32]
优化土壤调整植被指数(OSAVI)	$(N I R - R) / (N I R + R + 0.16)$	[33]
全色归一化植被指数(PNDVI)	$[N I R - (R + G + B)] / [N I R + (R + G + B)]$	[29]
红绿归一化植被指数(RBNDVI)	$[N I R - (R + B)] / [N I R + (R + B)]$	[29]
比值植被指数(RVI)	$N I R / R$	[25]
宽动态范围植被指数(WDRVI)	$(0.2 N I R - R) / (0.2 N I R + R)$	[34]

Tab.1

Fig.2

Tab.2

Fig.3

Fig.4

Tab.3

Tab.4

Fig.5

Fig.6

References 47

[1]	Lee K S, Kim D. Global dynamics of a pine wilt disease transmission model with nonlinear incidence rates[J]. Applied Mathematical Modelling, 2013, 37(6):4561-4569. doi: 10.1016/j.apm.2012.09.042
[2]	Manuel M M. Pine Wilt Disease:A Worldwide Threat to Forest Ecosystems[M]. Springer,Dordrecht: 2008.
[3]	叶建仁. 松材线虫病在中国的流行现状、防治技术与对策分析[J]. 林业科学, 2019, 55(9):1-10.
[4]	Syifa M, Park S J, Lee C W. Detection of the pine wilt disease tree candidates for drone remote sensing using artificial intelligence techniques[J]. Engineering, 2020, 6(8):919-926. doi: 10.1016/j.eng.2020.07.001
[5]	张竞成, 袁琳, 王纪华, 等. 作物病虫害遥感监测研究进展[J]. 农业工程学报, 2012, 28(20):1-11.
[6]	任艳中, 王弟, 李轶涛, 等. 无人机遥感在森林资源监测中的应用研究进展[J]. 中国农学通报, 2020, 36(8):111-118.
[7]	Iordache M D, Mantas V, Baltazar E, et al. A machine learning approach to detecting pine wilt disease using airborne spectral imagery[J]. Remote Sensing, 2020, 12(14):2280. doi: 10.3390/rs12142280
[8]	Dash J P, Watt M S, Pearse G D, et al. Assessing very high-resolution UAV imagery for monitoring forest health during a simulated disease outbreak[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2017, 131:1-14. doi: 10.1016/j.isprsjprs.2017.07.007
[9]	赵晋陵, 金玉, 叶回春, 等. 基于无人机多光谱影像的槟榔黄化病遥感监测[J]. 农业工程学报, 2020, 36(8):54-61.
[10]	Jaafar W S W M, Woodhouse I H, Silva C A, et al. Improving individual tree crown delineation and attributes estimation of tropical forests using airborne LiDAR data[J]. Forests, 2018, 9(12):759. doi: 10.3390/f9120759
[11]	Yu Run, Luo Youqing, Zhou Quan, et al. A machine learning algorithm to detect pine wilt disease using UAV-based hyperspectral imagery and LiDAR data at the tree level[J]. International Journal of Applied Earth Observation and Geoinformation, 2021, 101(1):102363 doi: 10.1016/j.jag.2021.102363
[12]	Navarro J A, Algeet N, Fernández-Landa A, et al. Integration of UAV, Sentinel-1,and Sentinel-2 data for mangrove plantationaboveground biomass monitoring in Senegal[J]. Remote Sensing, 2019, 11(1):77. doi: 10.3390/rs11010077
[13]	Ok A O, Ozdarici-Ok A. 2-D delineation of individual citrus trees from UAV-based dense photogrammetric surface models[J]. International Journal of Digital Earth, 2018, 11(6):583-608. doi: 10.1080/17538947.2017.1337820
[14]	Nevalainen O, Honkavaara E, Tuominen S, et al. Individual tree detection and classification with UAV-based photogrammetric point clouds and hyperspectral imaging[J]. Remote Sensing, 2017, 9(3):185. doi: 10.3390/rs9030185
[15]	Luís P, Pedro M, Luís M, et al. Monitoring of chestnut trees using machine learning techniques applied to UAV-Based multispectral data[J]. Remote Sensing, 2020, 12(18):3032. doi: 10.3390/rs12183032
[16]	Minařík, Langhammer J,Lendzioch T. Detection of bark beetle disturbance at tree level using UAS multispectral imagery and deep learning[J]. Remote Sensing, 2021, 13(23):4768. doi: 10.3390/rs13234768
[17]	Cardil A, Otsu K, Pla M, et al. Quantifying pine processionary moth defoliation in a pine-oak mixed forest using unmanned aerial systems and multispectral imagery.[J]. PLoS ONE, 2019, 14(3):e0213027-e0213027. doi: 10.1371/journal.pone.0213027
[18]	Minařík, Langhammer J,Lendzioch T. Automatic Tree Crown Extraction from UAS Multispectral Imagery for the Detection of Bark Beetle Disturbance in Mixed Forests[J]. Remote Sensing, 2020, 24(12):4081.
[19]	徐华潮, 骆有庆, 张廷廷, 等. 松材线虫自然侵染后松树不同感病阶段针叶光谱特征变化[J]. 光谱学与光谱分析, 2011, 31(5):1352-1356.
[20]	Santos CSS, Vasconcelos MW. Identification of genes differentially expressed in Pinus pinaster and Pinus pinea after infection with the pine wood nematode[J]. EUR J PLANT PATHOL, 2012, 132(3):407-418. doi: 10.1007/s10658-011-9886-z
[21]	Li Wenkai, Guo Qinghua, Jakubowski M K, et al. A new method for segmenting individual trees from the lidar point cloud[J]. Photogrammetric Engineering & Remote Sensing, 2012, 78(1):75-84.
[22]	Lu Xingcheng, Guo Qinghua, Li Wenkai, et al. A bottom-up approach to segment individual deciduous trees using leaf-off lidar point cloud data[J]. ISPRS Journal of Photogrammetry and Remote sensing, 2014, 94:1-12. doi: 10.1016/j.isprsjprs.2014.03.014
[23]	Gitelson A A, Kaufman Y J, Merzlyak M N. Use of a green channel in remote sensing of global vegetation from EOS-MODIS[J]. Remote sensing of Environment, 1996, 58(3):289-298. doi: 10.1016/S0034-4257(96)00072-7
[24]	Vincini M, Frazzi E, D’Alessio P, et al. A broad-band leaf chlorophyll vegetation index at the canopy scale[J]. PrecisionAgriculture, 2008, 9(5):303-319.
[25]	Jordan C F. Derivation of leaf-area index from quality of light on the forest floor[J]. Ecology, 1969, 50(4):663-666. doi: 10.2307/1936256
[26]	Ahamed T, Tian L, Zhang Y, et al. A review of remote sensing methods for biomass feedstock production[J]. Biomass and bioenergy, 2011, 35(7):2455-2469. doi: 10.1016/j.biombioe.2011.02.028
[27]	Miura T, Yoshioka H, Fujiwara K, et al. Inter-comparison of ASTER and MODIS surface reflectance and vegetation index products for synergistic applications to natural resource monitoring[J]. Sensors, 2008, 8(4):2480-2499. pmid: 27879830
[28]	Jiang Z, Huete A R, Didan K, et al. Development of a two-band enhanced vegetation index without a blue band[J]. Remote sensing of Environment, 2008, 112(10):3833-3845. doi: 10.1016/j.rse.2008.06.006
[29]	Wang Fumin, Huang Jingfeng, Tang Yanlin, et al. New vegetation index and its application in estimating leaf area index of rice[J]. Rice Science, 2007, 14(3):195-203.
[30]	Tucker C J, Elgin Jr J H, McMurtreyIii J E, et al. Monitoring corn and soybean crop development with hand-held radiometer spectral data[J]. Remote Sensing of Environment, 1979, 8(3):237-248. doi: 10.1016/0034-4257(79)90004-X
[31]	Qi J G, Chehbouni A R, Huete A R, et al. A modified soil adjusted vegetation index[J]. Remote Sensing of Environment, 1994, 48(2):119-126. doi: 10.1016/0034-4257(94)90134-1
[32]	Fitzgerald G J, Rodriguez D, Christensen L K, et al. Spectral and thermal sensing for nitrogen and water status in rainfed and irrigated wheat environments[J]. Precision Agriculture, 2006, 7(4):233-248. doi: 10.1007/s11119-006-9011-z
[33]	Steven M D. The sensitivity of the OSAVI vegetation index to observational parameters[J]. Remote Sensing of Environment, 1998, 63(1):49-60. doi: 10.1016/S0034-4257(97)00114-4
[34]	Gitelson A A. Wide dynamic range vegetation index for remote quantification of biophysical characteristics of vegetation[J]. Journal of plant physiology, 2004, 161(2):165-173. pmid: 15022830
[35]	刘笑笑, 王亮, 徐胜华, 等. 一种后向迭代的森林生物量遥感特征选择方法[J]. 测绘科学, 2017, 42(5):100-105.
[36]	Breiman L. Random forests[J]. Machine learning, 2001, 45(1):5-32. doi: 10.1023/A:1010933404324
[37]	VapnikV. The nature of statistical learning theory[M]. New York: Springer Verlag, 1995:25-27.
[38]	Snoek J, Larochelle H, Adams R P. Practical Bayesian optimization of machine learning algorithms[C]// Proceedings of the 25th International Conference on Neural Information Processing Systems-Volume 2.(2012-08-29)[2022-08-07].http://arXiv.org/abs/1206.2944.
[39]	Windrim L, Carnegie A J, Webster M, et al. Tree detection and health monitoring in multispectral aerial imagery and photogrammetric pointclouds using machine learning[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 13:2554-2572. doi: 10.1109/JSTARS.2020.2995391
[40]	金玉. 槟榔黄化病多源遥感数据监测研究[D]. 合肥: 安徽大学, 2020.
[41]	Xu Zhong, Shen Xin, Cao Lin, et al. Tree species classification using UAS-based digital aerial photogrammetry point clouds and multispectral imageries in subtropical natural forests[J]. International Journal of Applied Earth Observation and Geoinformation, 2020, 92:102173 doi: 10.1016/j.jag.2020.102173
[42]	Guerra-Hernández J, Cosenza D N, Rodriguez L C E, et al. Comparison of ALS-and UAV(SfM)-derived high-density point clouds for individual tree detection in Eucalyptus plantations[J]. International Journal of Remote Sensing, 2018, 39(15/16):5211-5235. doi: 10.1080/01431161.2018.1486519
[43]	刘见礼, 张志玉, 倪文俭, 等. 无人机影像匹配点云单木识别算法[J]. 遥感信息, 2019, 34(1):93-101.
[44]	曾健, 张晓丽, 周雪梅, 等. 倾斜摄影测量技术提取落叶松人工林地形信息[J]. 北京林业大学学报, 2019, 41(8):1-12.
[45]	刘家福, 李林峰, 任春颖, 等. 基于特征优选的随机森林模型的黄河口滨海湿地信息提取研究[J]. 湿地科学, 2018, 16(2):97-105.
[46]	刘文雅, 潘洁. 基于神经网络的马尾松叶绿素含量高光谱估算模型[J]. 应用生态学报, 2017, 28(4):1128-1136. doi: 10.13287/j.1001-9332.201704.035
[47]	Yu Run, Luo Youqing, Zhou Quan, et al., et al. Early detection of pine wilt disease using deep learning algorithms and UAV-based multispectral imagery[J]. Forest Ecology and Management, 2021, 497:119493. doi: 10.1016/j.foreco.2021.119493

研究区	样地	实际株数	正检 (TP)/株	误检 (FP)/株	漏检 (FN)/株	召回率 (r)/%	准确率 (p)/%	F得分 (F-score)/%
A1	P1	290	236	63	54	81.38	78.93	80.14
A1	P2	302	243	58	59	80.46	80.73	80.60
A2	P3	171	147	32	24	85.96	82.12	84.00
A2	P4	192	158	42	34	82.29	79.00	80.61
A3	P5	102	91	15	11	89.22	85.85	87.50
A3	P6	126	109	17	17	86.51	86.51	86.51
统计		1183	984	227	199	83.18	81.26	82.21

模型	参数	准确度/%
RF	n_estimators=61,max_depth=30	86.03
SVM	kernel=‘RBF’, C=2.95953,gamma=3.91795	80.00

模型
RF	健康	83.33	74.07	78.43	84.40	0.74
	早期	62.07	78.26	69.23
	中期	89.29	78.13	83.33
	晚期	93.33	94.92	94.12
SVM	健康	85.19	76.67	80.70	76.09	0.66
	早期	48.00	66.67	55.81
	中期	83.33	60.60	70.18
	晚期	80.65	87.72	84.03