Estimation of Potato Above-Ground Biomass Based on Hyperspectral Characteristic Parameters of UAV and Plant Height
LIU Yang1, 2, 3, 4, FENG Hai-kuan1, 3, 4*, HUANG Jue2, YANG Fu-qin5, WU Zhi-chao1, 3, 4, SUN Qian1, 2, 3, 4, YANG Gui-jun1, 3, 4
1. Key Laboratory of Quantitative Remote Sensing in Agriculture, Ministry of Agriculture, Beijing Research Center for Information Technology in Agriculture, Beijing 100097, China
2. College of Surveying Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, China
3. National Engineering Research Center for Information Technology in Agriculture, Beijing 100097, China
4. Beijing Engineering Research Center for Agriculture Internet of Things, Beijing 100097, China
5. College of Civil Engineering, Henan University of Engineering, Zhengzhou 451191, China
摘要： 地上生物量（above-ground biomass, AGB）是评价作物长势及其产量估测的重要指标，对指导农业管理具有重要的作用。因此，快速准确地获取生物量信息，对于监测马铃薯生长状况，提高产量具有重要的意义。于马铃薯现蕾期、块茎形成期、块茎增长期、淀粉积累期、成熟期获取成像高光谱影像、实测株高（heigh, H）、地上生物量和地面控制点（ground control point, GCP）的三维空间坐标。首先基于无人机高光谱灰度影像结合GCP生成试验田的DSM（digital surface model, DSM），利用DSM提取马铃薯的株高（Hdsm）；然后利用无人机高光谱影像计算一阶微分光谱、植被指数和绿边参数，进而分析高光谱特征参数（hyperspectral characteristic parameters, HCPs）和绿边参数(green edge parameters, GEPs)与马铃薯AGB的相关性，每个生育期筛选出相关性较高的前7个高光谱特征参数和最优绿边参数(optimal green edge parameters, OGEPs)；最后基于HCPs，HCPs加入OGEPs，HCPs加入OGEPs和Hdsm的组合利用偏最小二乘回归（partial least square regression, PLSR）和随机森林（random forest, RF）估算不同生育期的AGB。结果表明：（1）提取的Hdsm与实测株高H高度拟合（R2=0.84，RMSE=6.85 cm，NRMSE=15.67%）；（2）每个生育期得到的最优绿边参数不完全相同，现蕾期、块茎增长期和淀粉积累期OGEPs为Rsum，块茎形成期和成熟期OGEPs分别为Drmin和SDr；（3）与仅使用HCPs估算AGB相比，使用HCPs加入OGEPs，HCPs加入OGEPs和Hdsm在马铃薯不同生育期可以提高AGB估算精度，且以后者为自变量提高精度的幅度更大；（4）每个生育期利用PLSR和RF估算AGB的建模和验证R2从现蕾期到块茎增长期呈上升趋势，随后开始降低，整体上R2呈先上升后下降的趋势，通过PLSR方法构建的估算AGB模型效果优于RF方法，其中块茎增长期表现效果最好。因此，高光谱特征参数中结合最优绿边参数和株高，并使用PLSR方法可以改善马铃薯AGB的估算效果。
关键词：马铃薯；地上生物量；高光谱特征参数；绿边参数；株高
Abstract：Above-ground biomass (AGB) is an important index to evaluate crop growth and yield estimation, and plays an important role in guiding agricultural management. Therefore, the rapid and accurate acquisition of biomass information is of great significance for monitoring the growth status of potato and improving the yield. The hyperspectral images, measured plant height (H), above-ground biomass and three-dimensional coordinates of ground control point (GCP) were obtained in budding potato period, tuber formation period, tuber growth period, starch accumulation period and mature period. Firstly, based on UAV hyperspectral image and GCP to generate the DSM of the experimental field, the plant height (Hdsm) of potato was extracted by DSM. Then the first-order differential spectrum, vegetation index and green edge parameters are calculated using UAV hyperspectral images. Furthermore, the correlation between hyperspectral characteristic parameter (HCPs), green edge parameter (GEPs) and potato AGB was analyzed. The first seven hyperspectral characteristic parameters and the optimal green edge parameter (OGEPs) with good correlation with AGB were selected for each growth period. Finally, the AGB of different growth period was estimated by partial least square regression (PLSR) and random forest (RF) based on the combination of HCPs, HCPs and OGEPs, HCPs and OGEPs and Hdsm. The results show that: (1) the Hdsm is highly fitted to H (R2=0.84，RMSE=6.85 cm，NRMSE=15.67%). (2) The optimal green edge parameters obtained in each growth period are not completely the same. The OGEPs of the budding period, the tuber growth period and the starch accumulation period are Rsum, and the OGEPs of the tuber formation period and the mature period are Drmin and SDr, respectively. (3) Compared with HCPs, the accuracy of AGB estimation could be improved by adding OGEPs to HCPs, OGEPs and Hdsm to HCPs at different growth period of potato, and the latter improved the accuracy more greatly. (4) The R2 of AGB modeling and verification estimated by PLSR and RF showed an upward trend from budding period to tuber growth period and then began to decrease. On the whole, R2 decreased after increased. The estimation of AGB by PLSR is better than RF in each growth period，among which the AGB estimation of tuber growth period was the best. Therefore, the estimation accuracy of potato AGB can be improved by combining the OGEPs and plant height in HCPs and using PLSR method.
刘 杨，冯海宽，黄 珏，杨福芹，吴智超，孙 乾，杨贵军. 基于无人机高光谱特征参数和株高估算马铃薯地上生物量[J]. 光谱学与光谱分析, 2021, 41(03): 903-911.
LIU Yang, FENG Hai-kuan, HUANG Jue, YANG Fu-qin, WU Zhi-chao, SUN Qian, YANG Gui-jun. Estimation of Potato Above-Ground Biomass Based on Hyperspectral Characteristic Parameters of UAV and Plant Height. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2021, 41(03): 903-911.
[1] Bareth G, Aasen H, Bendig J, et al. Photogrammetrie-Fernerkundung-Geoinformation, 2015, 2015(1): 69.
[2] Li Xinchuan, Zhang Youjing, Luo Juhua, et al. International Journal of Applied Earth Observation and Geoinformation, 2015, 44: 104.
[3] Liu Bing, Asseng Senthold, Wang Anning, et al. Agricultural & Forest Meteorology, 2017, 247: 476.
[4] CHEN Peng, FENG Hai-kuan, LI Chang-chun, et al(陈 鹏, 冯海宽, 李长春, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2019, 35(11): 63.
[5] Atzberger C, Darvishzadeh R, Immitzer M, et al. International Journal of Applied Earth Observation and Geoinformation, 2015, 43: 19.
[6] DUAN Ding-ding, HE Ying-bin, LUO Shan-jun, et al(段丁丁, 何英彬, 罗善军, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2018, 38(10): 3215.
[7] Jin X, Liu S, Baret F, et al. Remote Sensing of Environment, 2017, 198: 105.
[8] Jibo Y, Guijun Y, Changchun L, et al. Remote Sensing, 2017, 9(7): 708.
[9] Nijat Kasim, SHI Qing-dong, WANG Jing-zhe, et al(尼加提·卡斯木, 师庆东, 王敬哲, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2017, 33(22): 208.
[10] QIN Zhan-fei, CHANG Qing-rui, XIE Bao-ni, et al(秦占飞, 常庆瑞, 谢宝妮, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2016, 32(23): 77.
[11] Tao Huilin, Feng Haikuan, Xu Liangji, et al. Sensors, 2020, 20(5): 1296.
[12] LIU Chang, YANG Gui-jun, LI Zhen-hai, et al(刘 畅, 杨贵军, 李振海, 等). Scientia Agricultura Sinica(中国农业科学), 2018, 51(16): 3060.
[13] Li Bo, Xu Xiangming, Zhang Li, et al. ISPRS Journal of Photogrammetry and Remote Sensing, 2019, 162: 161.
[14] WANG Bei-zhan, FENG Xiao, WEN Nuan, et al(王备战, 冯 晓, 温 暖, 等). Scientia Agricultura Sinica(中国农业科学), 2012, 45(15): 3049.
[15] 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.
[16] Jibo Y, Haikuan F, Guijun Y, et al. Remote Sensing, 2018, 10(2): 66.
[17] LIU Jia, WANG Li-min, YANG Fu-gang, et al(刘 佳, 王利民, 杨福刚, 等). Chinese Agricultural Science Bulletin(中国农学通报), 2019, 35(6): 143.
[18] NIU Qing-lin, FENG Hai-kuan, YANG Gun-jun, et al(牛庆林, 冯海宽, 杨贵军, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2018, 34(5): 73.
[19] 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 for Agricultural Machinery(农业机械学报), 2020, 51(1): 176.
