Research on Apple Yield Estimation Model Based on Unmanned Aerial Vehicle Remote Sensing

doi:10.13304/j.nykjdb.2024.0159

Abstract

Abstract:

The rapid and accurate estimation of apple tree yield based on UAV remote sensing technology is of great significance for precise orchard management and market planning. Using a DJI Phantom 4 RTK quadcopter drone equipped with a multispectral camera， multispectral images of apple trees during the flowering period， fruit setting period， fruit expansion period， and maturity period were obtained. The images were stitched using DJI Terra software， and 13 vegetation indices and 4 phenotypic features including tree height， east-west canopy width， north-south canopy width and average canopy width were extracted. Subsequently， different apple yield estimation models were constructed for different growth stages and variable combinations using random forest （RF）， back propagation neural network （BPNN） and support vector regression （SVR）. The results showed that， compared to considering single variables， combining phenotypic features with vegetation indices yielded better yield estimation results. Among the 4 growth stages， the frui expansion period was more suitable for apple yield estimation. Among the 3 regression algorithms， SVR achieved the best yield estimation performance. The optimal yield estimation model was the P+VI-SVR_S3 model based on phenotypic features and vegetation indices during the fruit expansion period， with a coefficient of determination （R₂） of 0.802 0， mean absolute error （MAE） of 15.024 2， and root mean square error （RMSE） of 18.510 7. Using vegetation indices and phenotypic features， apple yield could be estimated， and combining phenotypic features with vegetation indices could improve the accuracy of apple yield estimation. SVR performed better for apple yield estimation during the fruit expansion period.

Key words: UAV remote sensing, apple yield, inversion, machine learning

摘要：

基于无人机遥感技术快速、准确地估测苹果树产量对于果园的精准管理与市场规划具有重要意义。利用DJI Phantom 4 RTK四旋翼无人机搭载多光谱相机分别获取苹果树花期、幼果期、果实膨大和成熟期的多光谱影像，通过大疆智图（DJI Terra）软件对无人机影像进行拼接，并提取出13种植被指数和株高、东西冠幅、南北冠幅、平均冠幅4种表型特征，再分别利用随机森林（random forest，RF）模型、BP神经网络（back propagation neural network，BPNN）模型和支持向量回归（support vector regression，SVR）模型构建不同生育期、不同输入变量组合的苹果产量估测模型。结果表明，相对于单一变量，表型特征结合植被指数进行产量评估的效果更好。4个生育期中果实膨大期更利于苹果产量估测。3种回归算法中，利用SVR模型构建的产量估测模型效果最好。最佳产量估测模型为果实膨大期基于表型特征与植被指数结合的P+VI-SVR_S3模型，决定系数（R²）为0.802 0，平均绝对误差为15.024 2，均方根误差为18.510 7。综上，利用植被指数和表型特征相结合能够提升苹果产量估测精度，在果实膨大期利用SVR能更准确地估测苹果产量。

关键词: 无人机遥感, 苹果产量, 反演, 机器学习

CLC Number:

S126

Zhenfei ZHANG, An YAN, Jing GUO, Yuhang ZHAO, Yilin YUAN, Peng LIU, Zuohao QU, Chuan YUAN. Research on Apple Yield Estimation Model Based on Unmanned Aerial Vehicle Remote Sensing[J]. Journal of Agricultural Science and Technology, 2025, 27(9): 110-119.

张振飞, 颜安, 郭靖, 赵宇航, 袁以琳, 刘鹏, 曲佐昊, 袁川. 基于无人机遥感的苹果产量估测模型研究[J]. 中国农业科技导报, 2025, 27(9): 110-119.

Figures/Tables 6

Table 1 Vegetation indices and their calculation formulas

植被指数 Vegetation index	计算公式 Computational formulation	参考文献Reference
归一化差异植被指数 Normalized difference vegetation index（NDVI）	$N D V I = N I R - R N I R + R$	［26］
差值植被指数 Different influential factors（DVI）	$D V I = N I R - R$	［27］
比值植被指数 Ratio vegetation index（RVI）	$R V I = N I R R$	［28］
绿色归一化差异植被指数 Green normalized difference vegetation index（GNDVI）	$G N D V I = N I R - G N I R + G$	［29］
蓝色归一化差异植被指数 Blue normalized difference vegetation index（BNDVI）	$B N D V I = N I R - B N I R + B$	［30］
归一化差异红边植被指数 Normalized difference red edge index（NDRE）	$N D R E = N I R - R E N I R + R E$	［31］
红边叶绿素指数 Chlorophyll index-red edge（CIrededge）	$C I r e d e d g e = N I R R E - 1$	［32］
绿色叶绿素指数 Chlorophyll index-green（CIgreen）	$C I g r e e = N I R G - 1$	［32］
超绿指数 Excess green index（ExG）	$E x G = 2 G - R - B$	［33］
归一化差异指数 Normalized green-red difference index（NGRDI）	$N G R D I = G - R G + R$	［34］
可见光差异植被指数 Visible-band difference vegetation index（VDVI）	$V D V I = 2 G - R - B 2 G + R + B$	［35］
增强型植被指数 Enhanced vegetation index（EVI）	$E V I = 2.5 (N I R - R) N I R + 6 R - 7.5 B + 1$	［36］
土壤调节植被指数 Soil adjusted vegetable index（SAVI）	$S A V I = (1 + L) (N I R - R) N I R + R + L$	［37］

Table 1 Vegetation indices and their calculation formulas

植被指数 Vegetation index	计算公式 Computational formulation	参考文献Reference
归一化差异植被指数 Normalized difference vegetation index（NDVI）	$N D V I = N I R - R N I R + R$	［26］
差值植被指数 Different influential factors（DVI）	$D V I = N I R - R$	［27］
比值植被指数 Ratio vegetation index（RVI）	$R V I = N I R R$	［28］
绿色归一化差异植被指数 Green normalized difference vegetation index（GNDVI）	$G N D V I = N I R - G N I R + G$	［29］
蓝色归一化差异植被指数 Blue normalized difference vegetation index（BNDVI）	$B N D V I = N I R - B N I R + B$	［30］
归一化差异红边植被指数 Normalized difference red edge index（NDRE）	$N D R E = N I R - R E N I R + R E$	［31］
红边叶绿素指数 Chlorophyll index-red edge（CIrededge）	$C I r e d e d g e = N I R R E - 1$	［32］
绿色叶绿素指数 Chlorophyll index-green（CIgreen）	$C I g r e e = N I R G - 1$	［32］
超绿指数 Excess green index（ExG）	$E x G = 2 G - R - B$	［33］
归一化差异指数 Normalized green-red difference index（NGRDI）	$N G R D I = G - R G + R$	［34］
可见光差异植被指数 Visible-band difference vegetation index（VDVI）	$V D V I = 2 G - R - B 2 G + R + B$	［35］
增强型植被指数 Enhanced vegetation index（EVI）	$E V I = 2.5 (N I R - R) N I R + 6 R - 7.5 B + 1$	［36］
土壤调节植被指数 Soil adjusted vegetable index（SAVI）	$S A V I = (1 + L) (N I R - R) N I R + R + L$	［37］

Fig. 1 Correlation analysis between characteristics of different growth stage and yieldNote：* and ** indicate significant correlations at P<0.05 and P<0.01 levels， respectively.

Table. 2 Apple yield estimates

生育期Growing stage	输入变量Input variable	随机森林 RF			支持向量回归 SVR			BP神经网络 BPNN
生育期Growing stage	输入变量Input variable	决定系数 R²	平均绝对误差 MAE	均方根误差 RMSE	决定系数 R²	平均绝对误差 MAE	均方根误差 RMSE	决定系数 R²	平均绝对误差 MAE	均方根误差 RMSE
花期Flower stage	P	0.206 3	32.203 1	37.062 0	0.341 9	28.535 1	33.748 5	0.119 4	32.911 2	39.039 7
	VI	0.364 8	24.201 8	33.155 6	0.382 9	24.823 4	32.680 3	0.316 1	25.462 4	34.404 9
	P+VI	0.432 1	24.929 9	31.351 1	0.500 6	21.369 2	29.399 7	0.527 2	21.731 5	28.604 9
幼果期Fruit formation stage	P	0.252 3	31.709 3	35.972 7	0.410 7	28.752 2	31.934 9	0.314 9	29.581 6	34.434 7
	VI	0.446 0	25.514 6	30.964 3	0.473 0	25.048 6	30.200 9	0.367 2	27.054 2	33.094 6
	P+VI	0.481 1	24.964 2	29.966 8	0.628 5	20.424 5	25.355 2	0.565 7	21.845 0	27.414 8
果实膨大期Fruit expansion stage	P	0.329 2	29.798 4	34.072 0	0.356 9	29.590 6	33.362 3	0.337 6	29.686 2	33.857 9
	VI	0.635 9	20.149 9	25.103 6	0.632 7	20.472 2	25.213 7	0.638 0	20.394 9	25.031 9
	P+VI	0.704 7	18.574 0	22.608 6	0.802 0	15.024 2	18.510 7	0.781 3	16.015 7	19.456 8
成熟期Fruit ripening stage	P	0.342 5	28.582 9	33.732 7	0.419 7	26.972 4	31.692 2	0.381 9	27.377 4	32.706 8
	VI	0.567 9	22.257 9	27.345 3	0.594 3	21.446 9	26.497 2	0.566 1	21.922 5	27.403 0
	P+VI	0.582 3	21.705 4	26.887 4	0.633 8	20.016 0	25.174 4	0.600 8	21.519 5	26.284 6

Fig. 2 Analysis of apple yield estimation resultsA： Flower stage； B： Fruit formation stage； C： Fruit formation stage； D： Fruit ripening stage.

Fig. 3 R2 distribution of 3 model

Fig. 4 Apple yield predicted result of P+VI-SVRS3 model

References 52

[1]	张鹏飞,王爱玲,田时敏,等.苹果不同生育期水肥需求规律研究进展[J].黑龙江农业科学,2020(8):121-125.
	ZHANG P F, WANG A L, TIAN S M, et al.. Research progress in water and fertilizer demand rule of apple in different growth stages [J]. Heilongjiang Agric. Sci., 2020(8):121-125.
[2]	杨永娥,张晓煜,梁小娟,等.苹果花期至幼果期花朵和果实对低温的敏感性研究[J].西北植物学报,2023,43(10):1694-1703.
	YANG Y E, ZHANG X Y, LIANG X J, et al.. Studies on the sensitivity of apple flowers and fruits to low temperatures from flowering to young fruiting stage [J]. Acta Bot. Bor-Occid. Sin., 2023, 43(10):1694-1703.
[3]	马友福,卢磊,唐金.新疆苹果品种选育现状、存在的问题及对策[J].果树资源学报,2023,4(5):67-71, 79.
	MA Y F, LU L, TANG J. Currentstatus, problems and countermeasures of apple cultivar breeding in Xinjiang [J]. J. Fruit Res., 2023, 4(5):67-71, 79.
[4]	霍学喜,刘天军,刘军弟,等.2020年度中国苹果产业发展报告(精简版)[J].中国果菜,2022, 42(2):1-6.
[5]	CHEN R Q, ZHANG C J, XU B, et al.. Predicting individual apple tree yield using UAV multi-source remote sensing data and ensemble learning [J/OL]. Comput. Electron. Agric., 2022, 201:107275 [2024-02-10]. .
[6]	HE L L, FANG W T, ZHAO G N, et al.. Fruit yield prediction and estimation in orchards:a state-of-the-art comprehensive review for both direct and indirect methods [J/OL]. Comput. Electron. Agric., 2022, 195:106812 [2024-02-10]. .
[7]	HASSAN M A, YANG M J, RASHEED A, et al.. A rapid monitoring of NDVI across the wheat growth cycle for grain yield prediction using a multi-spectral UAV platform [J]. Plant Sci., 2019, 282:95-103.
[8]	TAO H, FENG H, XU L, et al.. Estimation of the yield and plant height of winter wheat using UAV-based hyperspectral images [J/OL]. Mol. Breed., 2020, 20(4):E1231 [2024-02-10]..
[9]	TANABE R, MATSUI T, TANAKA T S T. Winter wheat yield prediction using convolutional neural networks and UAV-based multispectral imagery [J/OL]. Field Crops Res., 2023, 291:108786 [2024-02-10]. .
[10]	FURUKAWA F, MARUYAMA K, SAITO Y K, et al.. Corn height estimation using UAV for yield prediction and crop monitoring [J/OL]. Appl. Agric. Environ., 2020, 22:5 [2024-02-10]. .
[11]	KUMAR C, MUBVUMBA P, HUANG Y, et al.. Multi-stage corn yield prediction using high-resolution uav multispectral data and machine learning models [J]. Agronomy, 2023, 13(5):1277-1286.
[12]	FENG A J, ZHOU J F, VORIES E D, et al.. Yield estimation in cotton using UAV-based multi-sensor imagery [J]. Biosyst. Eng., 2020, 193:101-114.
[13]	FENG A, ZHANG M, SUDDUTH K A, et al.. Cotton yield estimation from UAV-based plant height [J]. Trans. ASABE, 2019, 62(2):393-404.
[14]	SILVAE E DA, ROJO BAIO F H, RIBEIRO TEODORO L P, et al.. UAV-multispectral and vegetation indices in soybean grain yield prediction based on in situ observation [J/OL]. Remote. Sens. Appl. Soc. Environ., 2020, 18:100318 [2024-02-10]. .
[15]	赵晓庆,杨贵军,刘建刚,等.基于无人机载高光谱空间尺度优化的大豆育种产量估算[J].农业工程学报, 2017, 33(1):110-116.
	ZHAO X Q, YANG G J, LIU J G, et al.. Estimation of soybean breeding yield based on optimization of spatial scale of UAV hyperspectral image [J]. Trans. Chin. Soc. Agric. Eng., 2017, 33(1):110-116.
[16]	MAIMAITIJIANG M, SAGAN V, SIDIKE P, et al.. Soybean yield prediction from UAV using multimodal data fusion and deep learning [J/OL]. Remote Sens. Environ., 2020, 237:111599 [2024-02-10]. .
[17]	VIJAYAKUMA R V, COSTA L, AMPATZIDIS Y. Prediction of citrus yield with AI using ground-based fruit detection and UAV imagery [C]//Proceedings of the 2021 ASABE Annual International Virtual Meeting, 2021.
[18]	VAN BEEK J, TITS L, SOMERS B,et al..Temporal dependency of yield and quality estimation through spectral vegetation indices in pear orchards [J].Remote.Sens.,2015,7(8):9886-9903.
[19]	ORTENZI L, VIOLINO S, PALLOTTINO F, et al.. Early estimation of olive production from light drone orthophoto, through canopy radius [J/OL]. Drones, 2021, 5(4):118 [2024-02-10]. .
[20]	APOLO-APOLO O E, PÉREZ-RUIZ M, MARTÍNEZ-GUANTER J, et al.. A cloud-based environment for generating yield estimation maps from apple orchards using UAV imagery and a deep learning technique [J/OL].Mol.Metab.,2020,11:1086 [2024-02-10]. .
[21]	刘永波,高文波,何鹏,等.基于改进ResNet50模型的自然环境下苹果物候期识别[J].智慧农业,2023,5(2):13-22.
	LIU Y B, GAO W B, HE P, et al.. Apple phenological period identification in natural envi ronment based on improved ResNet50 model [J]. Smart Agric., 2023, 5(2):13-22.
[22]	陈日强,李长春,杨贵军,等.无人机机载激光雷达提取果树单木树冠信息[J].农业工程学报,2020,36(22):50-59.
	CHEN R Q, LI C C, ZHU Y J, et al.. Extraction of crown information from individual fruit tree by UAV LiDAR [J]. Trans. Chin. Soc. Agric. Eng., 2020, 36(22):50-59.
[23]	WANG P, FAN W Y, LI M Z, et al.. Extraction of individual tree height using a combination of large-scale aerial photo and LiDAR [J]. Adv. Mater. Res., 2011, 268-270:1157-1162.
[24]	束美艳,李世林,魏家玺,等.基于无人机平台的柑橘树冠信息提取[J].农业工程学报,2021,37(1):68-76.
	SHU M Y, LI S L, WEI J X,et al..Extraction of citrus crown parameters using UAV platform [J].Trans. Chin. Soc. Agric. Eng.,2021,37(1):68-76.
[25]	周辰琴,余拥军,方陆明,等.飞行高度与郁闭度对水杉冠幅提取影响的研究[J].林业资源管理,2022(1):150-156.
	ZHOU C Q, YU Y J, FANG L M, et al.. Effects of flight height and canopy density on canopy extraction of metasequoia glyptostroboides [J]. For. Resour. Manage., 2022(1):150-156.
[26]	CARLSON T N, RIPLEY D A.On the relation between NDVI,fractional vegetation cover,and leaf area index [J]. Remote. Sens. Environ., 1997, 62(3):241-252.
[27]	MOKARRAM M, HOJJATI M, ROSHAN G,et al..Modeling the behavior of vegetation indices in the salt dome of Korsia in North-East of Darab,Fars,Iran [J/OL].Model.Earth Syst.Environ., 2015,1(3):27 [2024-02-10]. .
[28]	JORDAN C F. Derivation of leaf-area index from quality of light on the forest floor [J]. Ecology, 1969, 50(4):663-666.
[29]	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. Sens. Environ., 1996, 58(3):289-298.
[30]	WANG J, DING J, ABULIMITI A, et al.. Quantitative estimation of soil salinity by means of different modeling methods and visible-near infrared (VIS-NIR) spectroscopy,Ebinur Lake Wetland,Northwest China [J/OL]. PeerJ, 2018, 6:e4703 [2024-02-10]. .
[31]	FITZGERALD G, RODRIGUEZ D, O’LEARY G. Measuring and predicting canopy nitrogen nutrition in wheat using a spectral index—the canopy chlorophyll content index (CCCI) [J]. Field Crops Res., 2010, 116(3):318-324.
[32]	黄侠,姜雪芹,张灿,等.红边波段在水稻生育期识别中的应用研究[J].测绘地理信息,2023,48(3):87-90.
	HUANG X, JIANG X Q, ZHANG C, et al.. Application of red edge band in identification of rice growth stages [J]. J. Geomat., 2023, 48(3):87-90.
[33]	WOEBBECKE D M, MEYER G E, VON BARGEN K, et al.. Color indices for weed identification under various soil, residue, and lighting conditions [J]. Trans. ASAE, 1995, 38(1):259-269.
[34]	GUERRERO J M, PAJARES G, MONTALVO M, et al.. Support vector machines for crop/weeds identification in maize fields [J]. Expert Syst. Appl., 2012, 39(12):11149-11155.
[35]	汪小钦,王苗苗,王绍强,等.基于可见光波段无人机遥感的植被信息提取[J].农业工程学报,2015,31(5):152-159.
	WANG X Q, WANG M M, WANG S Q, et al.. Extraction of vegetation information from visible unmanned aerial vehicle images [J]. Trans. Chin. Soc. Agric. Eng., 2015, 31(5):152-159.
[36]	PENG X, HAN W, AO J, et al.. Assimilation of LAI derived from UAV multispectral data into the SAFY model to estimate maize yield [J/OL]. Remote Sens., 2021, 13(6):1094 [2024-02-10]. .
[37]	ROUJEAN J L, BREON F M. Estimating PAR absorbed by vegetation from bidirectional reflectance measurements [J]. Remote. Sens. Environ., 1995, 51(3):375-384.
[38]	范军亮,王涵,廖振棋,等.基于纹理-颜色特征与植被指数融合的冬小麦LAI估测[J].农业机械学报,2023,54(7):347-359.
	FAN J L, WANG H, LIAO Z Q, et al.. Winter wheat leaf area index estimation based on texture-color features and vegetation indices [J]. Trans. Chin. Soc. Agric. Mach., 2023, 54(7):347-359.
[39]	宋荣杰,宁纪锋,常庆瑞,等.基于小波纹理和随机森林的猕猴桃果园遥感提取[J].农业机械学报,2018,49(4):222-231.
	SONG R J, NING J F, CHANG Q R, et al.. Kiwifruit orchard mapping based on wavelet textures and random forest [J]. Trans. Chin. Soc. Agric. Mach., 2018, 49(4):222-231.
[40]	CUI L G, TAO Y Q, DENG J, et al.. BBO-BPNN and AMPSO-BPNN for multiple-criteria inventory classification [J/OL]. Expert Syst. Appl., 2021, 175:114842 [2024-02-10]. .
[41]	SAMANTARAY S, SAHOO A. Prediction of runoff using BPNN, FFBPNN, CFBPNN algorithm in arid watershed: a case study [J]. Int. J. Knowl.-Based Intell. Eng. Syst., 2020, 24(3):243-251.
[42]	AWAD M, KHANNA R, AWAD M, et al.. Support vector machines for classification [J]. Eff. Learn. Mach. Theor. Concepts Appl. Eng. Syst. Des., 2015, 22:39-66.
[43]	CHANDRA M A, BEDI S S. Survey on SVM and their application in imageclassification [J]. Int. J. Inf. Technol., 2021, 13(5):1-11.
[44]	SHUJUAN D, CHUNLING L, TAO Z, et al.. Extraction of remote sensing information for lake salinity level based on SVM: a case from Badain Jaran desert in Inner Mongolia [J]. Remote Sens. Nat. Resour., 2016, 28(4):114-118.
[45]	梁栋,管青松,黄文江,等.基于支持向量机回归的冬小麦叶面积指数遥感反演[J].农业工程学报,2013,29(7):117-123.
	LIANG D, GUAN Q S, HUANG W J, et al.. Remote sensing inversion of leaf area index based on support vector machine regression in winter wheat [J]. Trans. Chin. Soc. Agric. Eng., 2013, 29(7):117-123.
[46]	BENDIG J, BOLTEN A, BENNERTZ S, et al.. Estimating biomass of barley using crop surface models (CSMs) derived from UAV-based RGB imaging [J]. Remote Sens., 2014, 6(11):10395-10412.
[47]	YUE J B, FENG H K, YANG G J, et al.. A comparison of regression techniques for estimation of above-ground winter wheat biomass using near-surface spectroscopy [J/OL]. Remote. Sens., 2018, 10:66 [2024-02-10]. .
[48]	SUN G, WANG X, YANG H, et al.. A canopy information measurement method for modern standardized apple orchards based on UAV multimodal information [J/OL]. Sensors (basel switz.), 2020, 20(10):E2985 [2024-02-10]. .
[49]	岳继博,杨贵军,冯海宽.基于随机森林算法的冬小麦生物量遥感估算模型对比[J].农业工程学报,2016,32(18):175-182.
	YUE J B, YANG G J, FENG H K. Comparative of remote sensing estimation models of winter wheat biomass based on random forest algorithm [J]. Trans. Chin. Soc. Agric. Eng., 2016, 32(18):175-182.
[50]	郭斌,吴亚颜,周嘉耀,等.基于无人机影像数据的苹果产量模拟:以洛川县为例[J].西安科技大学学报,2017,37(3):450-454.
	GUO B, WU Y Y, ZHOU J Y, et al.. Simulation of the apple yield based on unmanned aerial vehicle image data—taking Luochuan County as an example [J]. J. Xi’an Univ. Sci. Technol., 2017, 37(3):450-454.
[51]	胡艳萍.不同品种柑橘叶片和根中磷的季节性变化的生理机制[D].重庆:西南大学,2023.
	HU Y P. Physiological mechanism of seasonal changes of phosphorus in leaves and roots of different Citrus cultivars [D]. Chongqing: Southwest University, 2023.
[52]	梁晨欣.柑橘遥感识别方法研究:以广西武鸣为例[D].北京:中国农业科学院,2021.
	LIANG C X. Remote sensing classification for Citrus [D]. Beijing: Chinese Academy of Agricultural, 2021.