基于集成学习的水稻氮素营养及籽粒蛋白含量监测

doi:10.3964/j.issn.1000-0593(2022)06-1956-09

摘要
参考文献
相关文章 (15)

全文: PDF (5183 KB)
输出: BibTeX | EndNote (RIS)

摘要：利用高光谱遥感技术在水稻收获前对籽粒品质相关的蛋白质含量进行监测，一方面可以及时调整栽培管理方式，指导合理追肥，另一方面，有助于提前掌握籽粒品质信息，明确市场定位。该研究以广东省典型优质籼稻为研究目标，基于2019年和2020年两年氮肥梯度实验，以水稻分化期和抽穗期冠层尺度高光谱数据、水稻氮素参数，包括叶片氮素含量（LNC）、叶片氮素积累量（LNA）、植株氮素含量（PNC）、植株氮素积累量（PNA）及籽粒蛋白含量数据为基础，利用四种个体机器学习算法partial least square regression (PLSR)、K-nearest neighbor (KNN)、Bayesian ridge regression (BRR)、support vector regression (SVR)，三种集成学习算法random forest (RF)、adaboost、bagging，针对水稻不同生育期氮素状况进行监测建模，在此基础上构建基于水稻冠层光谱信息、光谱信息结合水稻农学氮素参数的籽粒蛋白含量的监测模型，并对模型进行精度对比。研究结果表明，在水稻氮素营养监测方面，利用水稻冠层454~950 nm波段信息，采用RF及Adaboost算法，在水稻分化期、抽穗期及全生育期LNC、LNA、PNC及PNA模型R²均达到0.90以上，同时也具有较低的RMSE和MAE。在水稻籽粒蛋白品质监测方面，采用全波段光谱信息进行籽粒蛋白含量监测时，RF具有最高的精确度与稳定性，两生育期的RF模型对籽粒蛋白含量的监测结果R²分别为0.935和0.941，RMSE分别为0.235和0.226，MAE分别为0.189和0.152；两生育期以全波段光谱信息结合长势参数进行籽粒蛋白监测时，Adaboost模型具有最高的精确度和稳定性，其中分化期全波段光谱信息结合PNA作为输入参数，Adaboost模型R²为0.960，RMSE为0.175，MAE为0.150，以抽穗期全波段光谱信息结合PNC作为输入参数，R²为0.963，RMSE为0.170，MAE为0.137。研究结果表明，与PLSR，KNN，BRR和SVR几种个体学习器算法相比，集成算法RF，Adaboost和Bagging具备良好的处理多重共线性的能力，适合用于高光谱数据的分析与处理，在作物氮素营养监测及水稻品质的早期遥感监测方面具有明显优势。

关键词：高光谱遥感；水稻品质；机器学习；集成算法；Adaboost；Random forest

Abstract：The use of hyperspectral remote sensing technology to monitor the protein content related to grain quality before rice matures is important. It can promptly adjust cultivation management methods and guide reasonable fertilization and help to grasp rice grain quality information in advance and clarify market positioning. This study took typical high-quality Indica rice in Guangdong Province as the research goal. Two-year nitrogen gradient experiments were carried on in 2019 and 2020. The canopy level hyperspectral data and rice nitrogen parameters, including leaf nitrogen content (LNC), leaf nitrogen accumulation (LNA), plant nitrogen content (PNC), and plant nitrogen accumulation (PNA), were collected at the rice panicle initiation stage and heading stage. Then, four individual machine learning algorithms, Partial Least Square Regression (PLSR), K-Nearest Neighbor (KNN), Bayesian Ridge Regression (BRR), Support Vector Regression (SVR), and three ensemble learning algorithms, Random forest (RF), Adaboost, Bagging were used for monitoring and modeling the nitrogen status of rice at different growth stages. After that, the rice grain protein content estimation models based on rice canopy spectral information, and spectral information combined with rice nitrogen parameterswere constructed by different machine learning algorithms. The rice nitrogen and grain protein content estimation models’ accuracy were evaluated and compared. The study results showed that for rice nitrogen nutrition monitoring, using the rice canopy spectral information from 454~950 nm, the R² of LNC, LNA, PNC and PNA estimation models based on RF and Adaboost algorithms achieved above 0.90 at the rice, heading stage, with low RMSE and MAE. Panicle initiation stage When using full-band spectral information to estimaterice grain protein content, RF had the highest accuracy and stability, with R² of 0.935 and 0.941, RMSE of 0.235 and 0.226, and MAE of 0.189 and 0.152 at rice panicle initiation and heading stage, respectively. Adaboost model has the highest accuracy and stability for seed protein monitoring with full-band spectral information combined with growth parameters at both fertility stages, at the panicle initiation stage, the Adaboost algorithm with full-band spectral and PNA data can reach the bestfor rice grain proteinestimation, the R², RMSE and MAE was 0.960, 0.175, and 0.150. While at heading stage, the R²,RMSE and MAE was 0.963, 0.170, 0.137,when using Adaboost algorithm with the full-band spectral and PNC data as input parameters. The results showed that the ensemble algorithms RF, Adaboost and Bagging have good ability to deal with multiple covariance compared with several individual learner algorithms PLSR, KNN, BRR and SVR.And they are suitable for the analysis and processing of hyperspectral data, which have obvious advantages in crop nitrogen nutrition monitoring and rice quality early monitoring through remote sensing.

Key words：Hyperspectral remote sensing; Rice grain protein; Machine Learning; Ensemble algorithms; Adaboost; Random forest

收稿日期: 2021-10-27 修订日期: 2022-02-25

中图分类号:

TP79

基金资助: 广东省重点领域研发计划项目（2019B020214002），国家自然科学基金项目（42171394）资助

通讯作者: 宋晓宇 E-mail: songxy@nercita.org.cn

作者简介: 张杰，1997年生，西安科技大学测绘科学与技术学院硕士研究生 e-mail: zhangj0216@163.com

引用本文:

张杰，徐波，冯海宽，竞霞，王娇娇，明世康，傅友强，宋晓宇. 基于集成学习的水稻氮素营养及籽粒蛋白含量监测[J]. 光谱学与光谱分析, 2022, 42(06): 1956-1964.
ZHANG Jie, XU Bo, FENG Hai-kuan, JING Xia, WANG Jiao-jiao, MING Shi-kang, FU You-qiang, SONG Xiao-yu. Monitoring Nitrogen Nutrition and Grain Protein Content of Rice Based on Ensemble Learning. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2022, 42(06): 1956-1964.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2022)06-1956-09 或 https://www.gpxygpfx.com/CN/Y2022/V42/I06/1956

[1] ZHANG Han, ZHAO Xiao-min, GUO Xi, et al(张晗, 赵小敏, 郭熙, 等). Jiangsu Agricultural Science(江苏农业科学), 2018, 46(12): 1.
[2] JIANG Huan-yu, YING Yi-bin, XIE Li-juan(蒋焕煜, 应义斌, 谢丽娟). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2008，28(6): 1300.
[3] TANG Yan-lin, WANG Ji-hua, HUANG Jing-feng, et al(唐延林, 王纪华, 黄敬峰, 等). Acta Agronomica Sinica（作物学报）, 2004，(8): 780.
[4] Liu X D, Sun Q H. International Journal of Pest Management, 2016, 62(3): 205.
[5] ZHANG Hao, HU Hao, CHEN Yi, et al(张浩, 胡昊, 陈义, 等). Journal of Nuclear Agricultural Sciences(核农学报), 2012, 26(1): 135.
[6] SUN Xue-mei, ZHOU Qi-fa, HE Qiu-xia(孙雪梅, 周启发, 何秋霞). Acta Agronomica Sinica（作物学报）, 2005，(7): 844.
[7] LIU Yun, TANG Yan-lin, HUANG Jing-feng, et al(刘芸, 唐延林, 黄敬峰, 等). Scientia Agricultura Sinica(中国农业科学), 2008, (9): 2617.
[8] Bennett K P, Parrado-Hernández E. The Journal of Machine Learning Research, 2006, 7: 1265.
[9] Liakos K G, Busato P, Moshou D, et al. Sensors, 2018，18(8): 2674.
[10] Bao Y, Mi C, Wu N, et al. Applied Sciences, 2019, 9(19): 4119.
[11] Chan J C W, Paelinckx D. Remote Sensing of Environment, 2008, 112(6): 2999.
[12] Gislason P O, Benediktsson J A, Sveinsson J R. Pattern Recognition Letters, 2006, 27(4): 294.
[13] Pham K, Kim D, Park S, et al. Catena, 2021, 196: 104886.
[14] ZHANG Fu-suo, CUI Zhen-ling, WANG Ji-qing, et al（张福锁, 崔振岭, 王激清, 等）. Chinese Bulletin Botany（植物学通报）, 2007，(6): 687.
[15] Fitzgerald M A, McCouch S R, Hall R D. Trends in Plant Science, 2009, 14(3): 133.
[16] ZHAO Xiao-min, SUN Xiao-xiang, WANG Fang-dong, et al（赵小敏, 孙小香, 王芳东, 等）. Acta Agriculturae Universitatis Jiangxiensis（江西农业大学学报）, 2019, 41(1): 1.
[17] Davari M, Karimi S A, Bahrami H A, et al. Catena, 2021, 197: 104987.
[18] Souza D V, Nievola J C, Santos J X, et al. Journal of Sustainable Forestry, 2019, 38(8): 755.
[19] Assaf A G, Tsionas M, Tasiopoulos A. Tourism Management, 2019, 71: 1.
[20] Chen Y W, Lin C J. Feature Extraction，2006, 207: 315.
[21] Belgiu M, Drǎguţ L. ISPRS Journal of Photogrammetry and Remote Sensing, 2016, 114: 24.
[22] Wu S, Nagahashi H. IEEE Signal Processing Letters, 2014, 21(6): 687.
[23] Wu X, Kumar V, Quinlan J R, et al. Knowledge and Information Systems, 2008, 14(1): 1.