Estimation of Aboveground Rice Biomass by Unmanned Aerial Vehicle Imaging
-
摘要:
目的 为探究无人机图像估算水稻地上部生物量(Aboveground biomass,AGB)的可行性,明确各图像特征与水稻AGB的定量关系,构建基于图像特征的水稻AGB估算模型。 方法 通过实施2个品种和4个施氮水平的小区试验,于分蘖期、孕穗期和齐穗期测定水稻AGB,同步采用无人机搭载数码相机获取水稻图像并提取颜色指数和纹理特征,分析其在不同生育期与水稻AGB之间的相关性,构建定量估算模型,并对模型进行检验。 结果 颜色指数中红蓝差值(r-b)与水稻AGB之间的相关性最好,纹理特征参数(G-mean)与水稻AGB之间的相关性最高;基于红蓝差值(r-b)和G-mean构建的水稻AGB双指数模型优于单一指数模型,全生育期估算模型y=2544.507+5054.243x1−145.543x2−556.553x1x2+27379.41x12+3.927x22,建模决定系数(R2)为0.9202,模型检验的决定系数(R2)为0.9112。 结论 基于颜色指数(r-b)和纹理特征参数(G-mean)融合构建的AGB估算模型可准确的估算水稻AGB,在水稻长势快速无损监测和精确管理中具有应用价值。 Abstract:Objective Feasibility of using images generated by unmanned aerial vehicle (UAV) to estimate the aboveground biomass (AGB) on a rice field was evaluated for crop production prediction. Methods On fields of two different varieties of rice fertilized with 4 varied nitrogen applications, AGB of rice plants at tillering, booting, and full heading stages were recorded by using the UAV imaging technology. Data on color and texture measurements were extracted from the images to correlate with corresponding AGB. A mathematic model was constructed, tested, and validated for prediction accuracy. Result On color, the red and blue differentiation (r-b) of the images highly correlated with the AGB; on texture, it was the G-mean. A prediction model was thus obtained for the entire growth period as y=2 544.507+5 054.243x1−145.543x2−556.553x1x2+27 379.41x12+3.927x22 , which had a correlation coefficient (R2) of 0.920 2 and a test determination coefficient of 0.911 2. Conclusion The prediction model based on r-b and G-mean derived from the UAV images performed satisfactorily in monitoring the AGB for the entire growth period of rice in the field. It was conceivably applicable for the farming operation and crop management. -
Key words:
- Rice /
- aboveground biomass /
- UAV image /
- estimation model
-
表 1 地上部生物量与颜色指数之间的相关性
Table 1. Correlation between AGB and color indices
序号
Number颜色指数
Color index相关系数r
Correlation coefficient分蘖期
Tillering孕穗期
Booting齐穗期
Heading全生育期
Whole growth period1 r-b −0.899** −0.914** −0.887** −0.890** 2 VARI 0.809** 0.844** 0.833** 0.841** 3 G −0.818** −0.834** −0.843** −0.841** 4 NRI −0.821** −0.833** −0.822** −0.823** 5 r/b −0.817** −0.825** −0.821** −0.822** 6 R −0.802** −0.821** −0.812** −0.811** 7 WI 0.792** 0.803** 0.793** 0.795** 8 NBI 0.764* 0.798** 0.792** 0.793** 9 MGRVI 0.748* 0.778* 0.772* 0.769* 10 GRVI 0.750* 0.757* 0.764* 0.766* 11 ExR −0.738* −0.745* −0.741* −0.746* 12 g/b −0.63* −0.73* −0.731* −0.728* 13 ExGR 0.532 0.614* 0.609* 0.611* 14 B −0.569 −0.583 −0.578 −0.572 15 g-b −0.478 −0.532 −0.523 −0.516 16 ExG 0.245 0.321 0.328 0.331 17 NGI 0.245 0.312 0.321 0.315 18 r+b −0.245 −0.311 −0.301 −0.312 19 RGBVI 0.042 0.052 0.049 0.051 *代表P=0.05显著水平,**代表P=0.01极显著水平。表5同。
Significance level: * represents P=0.05. Extremely significant level:** represents P=0.01.The same as Table 5.表 2 基于红蓝差值(r-b)的AGB估算模型构建和验证
Table 2. Construction and validation of AGB estimation model based on (r-b)
生育期
Growth stage建模
Calibration检验
Validation类型 Type 估算模型 Estimation model 决定系数R 2
Coefficient of determination R2RMSE/
(g·m−2)RRMSE/
%R2 分蘖期
Tillering二次多项式 Quadratic y = 2690.9x2 − 2806.5x + 942.59 0.8778 22.07 6.09 0.8682 指数 Exponential y = 972.12e−3.47x 0.8774 24.15 6.64 0.8639 对数 Logarithmic y = −355.2ln(x) − 83.638 0.8772 22.94 6.28 0.8568 乘幂 Power y = 100.57x−1.011 0.8748 24.83 6.78 0.8487 线性 Linear y = −1213.7x + 711.68 0.8719 24.25 6.65 0.8543 孕穗期
Booting二次多项式 Quadratic y = −1491.4x2 − 836.69x + 892.93 0.8820 20.22 5.58 0.8713 线性 Linear y = −1431.4x + 948.53 0.8801 20.54 5.66 0.8654 指数 Exponential y = 1017.8e−2.182x 0.8741 21.95 6.09 0.8501 对数 Logarithmic y = −268.9ln(x) + 220.54 0.8574 27.76 7.58 0.8435 乘幂 Power y = 336.6x−0.408 0.8409 29.73 8.09 0.8212 齐穗期
Heading二次多项式 Quadratic y = 3480.4x2 − 2549.8x + 1311.2 0.8728 23.99 6.62 0.8643 指数 Exponential y = 1292.7e−1.637x 0.8694 25.67 7.08 0.8566 线性 Linear y = −1778.5x + 1281.7 0.8630 26.46 7.25 0.8521 对数 Logarithmic y = −135ln(x) + 754.89 0.8159 27.9 7.68 0.8072 乘幂 Power y = 799.19x−0.123 0.7956 33.17 9.04 0.7013 全生育期
Whole growth period二次多项式 Quadratic y = 705.61x2 − 3450.2x + 1364.8 0.8693 24.65 6.74 0.8614 指数 Exponential y = 1688.3e−4.88x 0.8237 29.91 8.15 0.8122 线性 Linear y = −3177.4x + 1344.4 0.8689 26.59 7.30 0.8461 对数 Logarithmic y = −381.9ln(x) + 25.399 0.7251 35.06 9.62 0.6531 乘幂 Power y = 238.62x−0.547 0.5263 51.25 14.09 0.4101 表 3 地上部生物量与纹理特征参数之间的相关性
Table 3. Correlations between AGB and texture features
序号
Number纹理特征参数
Texture feature相关系数 r
correlation coefficient r分蘖期
Tillering孕穗期
Booting齐穗期
Heading全生育期
Whole growth period1 G-均值 −0.907** −0.932** −0.917** −0.901** G-mean 2 R-均值 −0.887** −0.891** −0.890** −0.889** R-mean 3 R-二阶矩 0.583* 0.782* 0.727* 0.801** R-second moment 4 B-均值 −0.768* −0.749* −0.762* −0.787* B-mean 5 R-方差 0.652* 0.742* 0.772* 0.713* R-variance 6 B-二阶矩 0.543 0.652* 0.544 0.522 B-second moment 7 G-方差 0.568 0.563 0.612* 0.555 G-variance 8 B-方差 0.464 0.494 0.432 0.564 B-variance 9 G-对比度 0.333 0.452 0.411 0.562 G-contrast 10 R-相异性 0.488 0.521 0.517 0.553 R-dissimilarity 11 B-信息熵 −0.327 −0.431 −0.417 −0.476 B-entropy 12 G-协同性 −0.415 −0.463 −0.434 −0.408 G-homogeneity 13 R-信息熵 −0.457 −0.325 −0.402 −0.376 R-entropy 14 G-相异性 0.404 0.441 0.454 0.412 G-dissimilarity 15 B-协同性 0.212 0.221 0.322 0.435 B-homogeneity 16 B-对比度 0.126 0.321 0.311 0.411 B-contrast 17 G-信息熵 −0.371 −0.362 −0.322 −0.377 G-entropy 18 R-协同性 0.195 0.211 0.322 0.164 R-homogeneity 19 B-相关性 −0.204 −0.264 −0.228 −0.224 B-correlation 20 G-二阶矩 0.04 0.211 0.142 0.226 G-second moment 21 R-相关性 −0.107 −0.124 −0.214 −0.135 R-correlation 22 R-对比度 0.118 0.123 0.135 0.074 R-contrast 23 B-相异性 0.093 0.132 0.127 0.211 B-dissimilarity 24 G-相关性 −0.006 −0.013 −0.021 −0.089 G-correlation 表 4 基于纹理特征参数(G-mean)的水稻各生育期AGB模型的构建和检验
Table 4. Construction and validation of AGB estimation model for each growth stage of rice plants based on texture G-mean
生育期
Growth stage建模 Calibration 检验 Validation 类型
Type估算模型
Estimation model决定系数R 2
Coefficient of
determination R2均方根误差
RMSE/
(g·m−2)相对均方根误差
RRMSE/
%决定系数
R2分蘖期
Tillering二次多项式 Quadratic y = 0.0736x2 − 14.46x + 780.83 0.8875 22.06 6.08 0.8797 对数 Logarithmic y = −327ln(x) + 1525.7 0.8873 21.67 6.01 0.8635 指数 Exponential y = 902.39e−0.026x 0.8871 23.89 6.56 0.8717 线性 Linear y = −9.0632x + 684.31 0.8859 23.65 6.48 0.8790 乘幂 Power y = 9934.4x−0.934 0.8742 24.45 6.70 0.8311 孕穗期
Booting二次多项式 Quadratic y = 0.1879x2 − 25.845x + 1201.6 0.9128 17.14 4.75 0.9034 指数 Exponential y = 1237.7e−0.024x 0.9126 16.63 4.63 0.9017 对数 Logarithmic y = −411.4ln(x) + 1994.6 0.9118 15.70 4.38 0.9004 线性 Linear y = −15.965x + 1075.6 0.9106 17.97 4.98 0.9094 乘幂 Power y = 5007.2x−0.627 0.9071 19.69 5.42 0.8903 齐穗期
Heading二次多项式 Quadratic y = 2.2052x2 − 114.72x + 2398.5 0.9014 17.26 4.81 0.8945 乘幂 Power y = 5702.9x−0.584 0.9001 17.52 4.95 0.8809 对数 Logarithmic y = −634.1ln(x) + 2892.2 0.8948 17.66 4.91 0.8843 指数 Exponential y = 1930.3e−0.033x 0.8888 20.01 5.57 0.8671 线性 Linear y = −35.866x + 1714.6 0.8778 22.64 6.22 0.8121 全生育期
Whole growth period二次多项式 Quadratic y = 0.9499x2 − 86.442x + 2254 0.9005 11.10 3.11 0.8993 乘幂 Power y = 45897x−1.33 0.8925 16.09 4.53 0.8713 对数 Logarithmic y = −865.4ln(x) + 3492.7 0.8913 13.95 3.89 0.8671 指数 Exponential y = 2447.4e−0.051x 0.8883 11.69 3.29 0.9098 线性 Linear y = −32.075x + 1557.3 0.8692 26.59 7.31 0.8463 表 5 基于r-b和G-mean的AGB双指数线性回归模型的构建和检验
Table 5. Construction and validation of AGB estimation model based on r-b and G-mean
生育期
Growth stage建模 Calibration 检验 Validation 估算模型
Model决定系数R 2
Coefficient of
determination R2均方根误差
RMSE/
(g·m−2)相对均方根误差
RRMSE/
%决定
系数
R2分蘖期 Tillering y=879.0997−3088.61x1+5.9053x2+364.8907x1x2−18580.4 x12−1.607 x22 0.9099 16.46 4.68 0.8903 孕穗期 Booting y=1218.506−1364.43x1−16.1009x2+380.0562x1x2−21120.2947x12−1.4778 x22 0.9155 22.01 3.32 0.9006 齐穗期 Heading y=2422.079+3430.592x1−143.099 x2−381.908x1x2+13707.691x12+4.369x22 0.9043 36.46 3.37 0.8901 全生育期
Whole growth periody=2544.507+5054.243x1−145.543x2−556.553x1x2+27379.41x12+3.927x22 0.9202 76.20 10.91 0.9112 -
[1] ZHU W, SUN Z, PENG J, et al. Estimating maize above-ground biomass using 3d point clouds of multi-source unmanned aerial vehicle data at multi-spatial scales [J]. Remote Sens, 2019, 11: 2678. doi: 10.3390/rs11222678 [2] 田婷, 张青, 张海东, 等. 基于无人机遥感的水稻产量估测 [J]. 中国稻米, 2022, 28(1):67−71, 77.TIAN T, ZHANG Q, ZHANG H D. Estimating yield of rice based on remote sensing by unmanned aerial vehicle [J]. China Rice, 2022, 28(1): 67−71, 77.(in Chinese) [3] 周蓉, 赵天忠, 吴发云. 基于Landsat 8遥感影像的地上生物量模型反演研究 [J]. 西北林学院学报, 2022, 37(2):186−192. doi: 10.3969/j.issn.1001-7461.2022.02.26ZHOU R, ZHAO T Z, WU F Y. Aboveground biomass model based on Landsat 8 remote sensing images [J]. Journal of Northwest Forestry University, 2022, 37(2): 186−192.(in Chinese) doi: 10.3969/j.issn.1001-7461.2022.02.26 [4] 姜妍, 王琳, 杨月, 等. 无人机高光谱成像技术在作物生长信息监测中的应用 [J]. 东北农业大学学报, 2022, 53(3):88−96.JIANG Y, WANG L, YANG Y, et al. Application of UAV hyperspectral imaging technology in crop growth information monitoring [J]. Journal of Northeast Agricultural University, 2022, 53(3): 88−96.(in Chinese) [5] 杨福芹, 李天驰, 冯海宽, 等. 基于无人机数码影像的冬小麦氮素营养诊断研究 [J]. 福建农业学报, 2021, 36(3):369−378.YANG F Q, LI T C, FENG H K, et al. UAV digital image-assisted monitoring on nitrogen nutrition of winter wheat in the field [J]. Fujian Journalof Agricultural Sciences, 2021, 36(3): 369−378.(in Chinese) [6] 梁胤豪, 陈全, 董彩霞, 等. 基于深度学习和无人机遥感技术的玉米雄穗检测研究 [J]. 福建农业学报, 2020, 35(4):456−464.LIANG Y H, CHEN Q, DONG C X, et al. Application of deep-learning and UAV for field surveying corn tassel [J]. Fujian Journal of Agricultural Sciences, 2020, 35(4): 456−464.(in Chinese) [7] YANG G J, LIU J G, ZHAO C J, et al. Unmanned aerial vehicle remote sensing for field-based crop phenotyping: Current status and perspectives [J]. Frontiers in Plant Science, 2017, 8: 1111. doi: 10.3389/fpls.2017.01111 [8] CASANOVA D, EPEMA G F, GOUDRIAAN J. Monitoring rice reflectance at field level for estimating biomass and LAI [J]. Field Crops Research, 1998, 55(1/2): 83−92. [9] TAKAHASHI W, NGUYEN-CONG V, KAWAGUCHI S, et al. Statistical models for prediction of dry weight and nitrogen accumulation based on visible and near-infrared hyper-spectral reflectance of rice canopies [J]. Plant Production Science, 2000, 3(4): 377−386. doi: 10.1626/pps.3.377 [10] GNYP M L, MIAO Y X, YUAN F, et al. Hyperspectral canopy sensing of paddy rice aboveground biomass at different growth stages [J]. Field Crops Research, 2014, 155: 42−55. doi: 10.1016/j.fcr.2013.09.023 [11] 王秀珍, 黄敬峰, 李云梅, 等. 水稻地上鲜生物量的高光谱遥感估算模型研究 [J]. 作物学报, 2003, 29(6):815−821. doi: 10.3321/j.issn:0496-3490.2003.06.004WANG X Z, HUANG J F, LI Y M, et al. Study on hyperspectral remote sensing estimation models for the ground fresh biomass of rice [J]. Acta Agronomica Sinica, 2003, 29(6): 815−821.(in Chinese) doi: 10.3321/j.issn:0496-3490.2003.06.004 [12] DAI M, YANG T L, YAO Z S, et al. Wheat biomass estimation in different growth stages based on color and texture features of UAV images [J/OL]. Smart Agriculture, Published online: 2022-02-22 14: 21: 52. URL: https://kns.cnki.net/kcms/detail/10.1681.S.20220221.1425.002.html. [13] 杨俊, 丁峰, 陈晨, 等. 小麦生物量及产量与无人机图像特征参数的相关性 [J]. 农业工程学报, 2019, 35(23):104−110. doi: 10.11975/j.issn.1002-6819.2019.23.013YANG J, DING F, CHEN C, et al. Correlation of wheat biomass and yield with UAV image characteristic parameters [J]. Transactions of the Chinese Society of Agricultural Engineering, 2019, 35(23): 104−110.(in Chinese) doi: 10.11975/j.issn.1002-6819.2019.23.013 [14] 刘畅, 杨贵军, 李振海, 等. 融合无人机光谱信息与纹理信息的冬小麦生物量估测 [J]. 中国农业科学, 2018, 51(16):3060−3073. doi: 10.3864/j.issn.0578-1752.2018.16.003LIU C, YANG G J, LI Z H, et al. Biomass estimation in winter wheat by UAV spectral information and texture information fusion [J]. Scientia Agricultura Sinica, 2018, 51(16): 3060−3073.(in Chinese) doi: 10.3864/j.issn.0578-1752.2018.16.003 [15] 陈晨. 基于无人机图像的小麦生物量与产量的估测研究[D]. 扬州: 扬州大学, 2019CHEN C. Estimation of wheat biomass and yield based on UAV images[D]. Yangzhou: Yangzhou University, 2019. (in Chinese) [16] 张建, 谢田晋, 尉晓楠, 等. 无人机多角度成像方式的饲料油菜生物量估算研究 [J]. 作物学报, 2021, 47(9):1816−1823.ZHANG J, XIE T J, WEI X N, et al. Estimation of feed rapeseed biomass based on multi-angle oblique imaging technique of unmanned aerial vehicle [J]. Acta Agronomica Sinica, 2021, 47(9): 1816−1823.(in Chinese) [17] 刘杨, 冯海宽, 孙乾, 等. 不同分辨率无人机数码影像的马铃薯地上生物量估算研究 [J]. 光谱学与光谱分析, 2021, 41(5):1470−1476.LIU Y, FENG H K, SUN Q, et al. Estimation study of above ground biomass in potato based on UAV digital images with different resolutions [J]. Spectroscopy And Spectral Analysis, 2021, 41(5): 1470−1476.(in Chinese) [18] 陶惠林, 徐良骥, 冯海宽, 等. 基于无人机数码影像的冬小麦株高和生物量估算 [J]. 农业工程学报, 2019, 35(19):107−116. doi: 10.11975/j.issn.1002-6819.2019.19.013TAO H L, XU L J, FENG H K, et al. Estimation of plant height and biomass of winter wheat based on UAV digital image [J]. Transactions of the Chinese Society of Agricultural Engineering, 2019, 35(19): 107−116.(in Chinese) doi: 10.11975/j.issn.1002-6819.2019.19.013 [19] 李艳大, 叶春, 曹中盛, 等. 基于作物生长监测诊断仪的双季稻叶片氮含量和氮积累量监测 [J]. 应用生态学报, 2020, 31(9):3040−3050. doi: 10.13287/j.1001-9332.202009.012LI Y D, YE C, CAO Z S, et al. Monitoring leaf nitrogen concentration and nitrogen accumulation of double cropping rice based on crop growth monitoring and diagnosis apparatus [J]. Chinese Journal of Applied Ecology, 2020, 31(9): 3040−3050.(in Chinese) doi: 10.13287/j.1001-9332.202009.012 [20] 叶春, 刘莹, 李艳大, 等. 基于RGB颜色空间的早稻氮素营养监测研究 [J]. 中国农业大学学报, 2020, 25(8):25−34. doi: 10.11841/j.issn.1007-4333.2020.08.03YE C, LIU Y, LI Y D, et al. Monitoring the nitrogen nutrition of early rice based on RGB color space [J]. Journal of China Agricultural University, 2020, 25(8): 25−34.(in Chinese) doi: 10.11841/j.issn.1007-4333.2020.08.03 [21] 石媛媛. 基于数字图像的水稻氮磷钾营养诊断与建模研究[D]. 杭州: 浙江大学, 2011SHI Y Y. Rice nutrition diagnosis and modeling based on digital image[D]. Hangzhou: Zhejiang University, 2011. (in Chinese) [22] 吴华兵, 朱艳, 田永超, 等. 棉花冠层高光谱参数与叶片氮含量的定量关系 [J]. 植物生态学报, 2007, 31(5):903−909. doi: 10.17521/cjpe.2007.0114WU H B, ZHU Y, TIAN Y C, et al. Relationship between canopy hyperspectra parameter and leaf nitrogen concentration in cotton [J]. Journal of Plant Ecology, 2007, 31(5): 903−909.(in Chinese) doi: 10.17521/cjpe.2007.0114 [23] LEE K J, LEE B W. Estimation of rice growth and nitrogen nutrition status using color digital camera image analysis [J]. European Journal of Agronomy, 2013, 48: 57−65. doi: 10.1016/j.eja.2013.02.011 [24] 班松涛. 水稻长势无人机遥感监测研究[D]. 杨凌: 西北农林科技大学, 2020BAN S T. Monitoring rice growth conditions using UAV remote sensing system[D]. Yangling: Northwest A&F University, 2020. (in Chinese)