文献类型： 外文期刊

作者： Bai, Yi ¹ ; Shi, Liangsheng ² ; Zha, Yuanyuan ² ; Liu, Shuaibing ¹ ; Nie, Chenwei ¹ ; Xu, Honggen ³ ; Yang, Hongye ³ ; Shao, Mingchao ¹ ; Yu, Xun ¹ ; Cheng, Minghan ¹ ; Liu, Yadong ¹ ; Lin, Tao ⁴ ; Cui, Ningbo ⁵ ; Wu, Wenbin ⁶ ; Jin, Xiuliang ¹ ;

作者机构： 1.Chinese Acad Agr Sci, Natl Nanfan Res Inst Sanya, Sanya 572024, Peoples R China

2.Wuhan Univ, State Key Lab Water Resources Hydropower Engn Sci, Wuhan 430072, Peoples R China

3.Chinese Acad Agr Sci, Key Lab Crop Physiol & Ecol, Minist Agr, Inst Crop Sci, Beijing 100081, Peoples R China

4.Xinjiang Acad Agr Sci, Inst Cash Crops, Urumqi 830091, Peoples R China

5.Sichuan Univ, Coll Water Resource & Hydropower, Lab Hydraul & Mt River Engn, Chengdu 610065, Peoples R China

6.Chinese Acad Agr Sci, Inst Agr Resources & Reg Planning, Beijing 100081, Peoples R China

关键词： Maize; Leaf age; UAV; Machine learning model; RGB and Multispectral image

期刊名称：COMPUTERS AND ELECTRONICS IN AGRICULTURE （ 影响因子：8.3； 五年影响因子：8.3 ）

ISSN： 0168-1699

年卷期： 2023 年 215 卷

页码：

收录情况： SCI

摘要： Leaf age is an essential parameter for describing the growth stage of crop. Thus, agronomists can develop timely cultivation strategies to promote maize growth according to leaf age. However, the traditional leaf age observation approach, which is inefficient, necessitates a large amount of people to investigate it in the field with destructive sampling. So, accurately quantifying leaf age under various maize types and production conditions remains a challenge. The purpose of this study was to develop a remote sensing monitoring approach for rapidly and non-destructively estimating the leaf age of maize seedlings. The UAV (unmanned aerial vehicle) highthroughput phenotyping platform was constructed to collect multi-source remote sensing images from maize emergence to jointing stage. Based on RGB and multispectral (MS) images, the image features of maize seedlings were extracted to construct the leaf age estimation models. The results showed that two regression models provided a reliable estimate performance of seedling leaf age, GBDT of which the best estimates are R2 of 0.88, Root Mean Square Error (RMSE) of 0.33, similarly, XGBoost being R2 of 0.89, RMSE of 0.32. The RGB-based model presented more accurate estimates (in terms of relative Root Mean Square Error of 9.26%) than the MS-based model (13.97%) and the RGB + MS-based model (12.26%). The results indicated that the maize seedling leaf age estimation method constructed in this study provides powerful technical support for agronomists to observe leaf age in the field.