Monitoring Potato Biomass and Plant Nitrogen Content With UAV-Based Hyperspectral Imaging
YANG Fu-qin1, CHEN Ri-qiang2, LIU Yang2, CHEN Xin-xin1, XIAO Yi-bo1, LI Chang-hao1, WANG Ping3, FENG Hai-kuan2, 4*
1. College of Civil Engineering, Henan University of Engineering, Zhengzhou 451191, China
2. Key Laboratory of Quantitative Remote Sensing in Agriculture of Ministry of Agriculture and Rural Affairs, Information Technology Research Center, Beijing Academy of Agriculture and Forestry Sciences, Beijing 100097, China
3. Hebei Zhongbao Green Crop Technology Corporation, Langfang 065000, China
4. National Engineering and Technology Center for Information Agriculture, Nanjing Agricultural University, Nanjing 210095, China
Abstract：Above-ground biomass and plant nitrogen content play a crucial role in crop growth, development, and yield formation. Therefore, dynamic monitoring of crop growth and nutritional status is of considerable importance. The study used unmanned aerial vehicles to obtain hyperspectral data and above-ground biomass during the budding stage, tuber formation stage, tuber growth and starch accumulation stage, to analyze the correlation and the importance of variable projection between vegetation indices and biomass and plant nitrogen content, and to screen out vegetation indices that are sensitive to biomass and plant nitrogen content combining deep neural network (DNN), partial least squares (PLSR), elastic network regression (ENR), ridge regression (RR) and support vector machine (SVR) to estimate biomass and plant nitrogen content and comparing the effectiveness of different models in estimating biomass and plant nitrogen content. The results showed that (1) the correlation between vegetation indices and both biomass and plant nitrogen content reached 0.01 significant level, and the importance of the variable projection was used to screen out the vegetation indices that were sensitive to biomass and plant nitrogen content; (2) Comparing the remote sensing estimation models for the five growth stages, the best model for biomass and plant nitrogen content was constructed at the tuber formation stage, the worst model for biomass was estimated at the present bud stage, and the worst model for plant nitrogen content was estimated at the tuber growth stage. (3) The optimum biomass model constructed in the tuber formation stage using the PLSR method was modelled with R2, RMSE and NRMSE was 0.60, 235.65 kg·hm-2 and 0.15 kg·hm-2 respectively, and validated with R2, RMSE and NRMSE was 0.58，344.72 kg·hm-2 and 0.26 kg·hm-2，The optimum plant nitrogen content model constructed during tuber formation stage using RR method was modelled with R2，RMSE and NRMSE was 0.74, 0.31% and 0.15%，validated R2, RMSE and NRMSE was 0.77, 0.58% and 0.28%. Comprehensively comparing the DNN, PLSR, ENR, RR, and SVR algorithms for estimating biomass and plant nitrogen content models, the accuracy of the estimated plant nitrogen content model is found to be better than that of the estimated biomass model. The plant's nitrogen content can be used to more effectively monitor crop growth and nutritional characteristics, providing a reference for informed agricultural management.
[1] LIU Yang, SUN Qian, HUANG Jue, et al(刘 杨，孙 乾，黄 珏，等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2021, 41(8): 2549.
[2] TAO Hui-lin, FENG Hai-kuan, YANG Gui-jun, et al(陶惠林, 冯海宽, 杨贵军, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2019, 35(23): 111.
[3] TAO Hui-lin, XU Liang-ji, FENG Hai-kuan, et al(陶惠林, 徐良骥, 冯海宽, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2019, 35(19): 107.
[4] Gee C, Denimal E. Remote Sensing, 2020, 12: 2982.
[5] Liu Y, Feng H K, Yue J B, et al. Computers and Electronics in Agriculture, 2022, 198: 107089.
[6] SHU Shi-fu, LI Yan-da, CAO Zhong-sheng, et al(舒时富, 李艳大, 曹中盛, 等). Fujian Journal of Agricultural Sciences(福建农业学报), 2022, 37(7)：824.
[7] LIU Yang, FENG Hai-kuan, HUANG Jue, et al(刘 杨, 冯海宽, 黄 珏, 等). Transactions of the Chinese Society for Agricultural Machinery(农业机械学报), 2021，52(2)：188.
[8] Wang Z L, Ma Y M, Chen Ping, et al. Frontiers in Plant Science，2022，13: 903643.
[9] WEI Peng-fei, XU Xin-gang, LI Zhong-yuan, et al(魏鹏飞, 徐新刚, 李中元, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2019, 35(8)：126.
[10] GUO Yan, JING Yu-hang, WANG Lai-gang, et al(郭 燕, 井宇航, 王来刚, 等). Scientia Agricultura Sinica(中国农业科学), 2023, 56（5）: 850.
[11] Yi Q X, Huang J F, Wang F M, et al. Environmental Science & Technology，2007，41(19) ：6770.
[12] Liu Y, Feng H, Yue J. Remote Sensing, 2022, 14: 5449.
[13] Vogelmann J E, Rock B N, Moss G D M. International Journal of Romote Sensing, 1993, 14(8): 1563.
[14] Blackburn G A. International Journal of Romote Sensing, 1998, 19: 657.
[15] Qi J, Chehbouni A, Huete A R, et al. Remote Sensing of Environment, 1994, 48(2): 119.
[16] Gitelson A A, Merzlyak M N. International Journal of Romote Sensing, 1997, 18: 2691.
[17] Gitelson A A, Vina V, Arkebauer T J, et al. Geophysical Research Letters, 2003, 30(5): 1248.
[18] Datt B. Journal of Plant Physiology, 1999, 154(1): 30.
[19] Rama Rao N, Garg P K, Ghosh S K, et al. Journal of Agricultural Science, 2008, 146(1): 65.
[20] Jin X L, Li Z H, Fang H K, et al. The Crop Journal, 2020, 8(1): 87.
[21] Fei S P, Chen Z, Li L, et al. Agricultural and Forest Meteorology, 2023, 328: 109237.
