Prediction of wheat SPAD using integrated multispectral and support vector machines

• DOI: 10.3389/fpls.2024.1405068
• Publication Date: 2024-06-20T14:10:23Z
• Authors: Wei Wang; Na Sun; Bin Bai; Hao Wu; Yukun Cheng; Hongwei Geng; JiKun Song; JinPing Zhou; Zhiyuan Pang; SongTing Qian; Wanyin Zeng
• Keywords: SPAD; multispectral; SVM; UAV; Plant culture; 0401 agriculture, forestry, and fisheries; Plant Science; 04 agricultural and veterinary sciences; winter wheat; SB1-1110

Rapidly obtaining the chlorophyll content of crop leaves is great significance for timely diagnosis health and effective field management. Multispectral imagery obtained from unmanned aerial vehicles (UAV) being used to remotely sense SPAD (Soil Plant Analyzer Development) values wheat crops. However, existing research has not yet fully considered impact different growth stages populations on accuracy estimation. In this study, 300 materials winter natural in Xinjiang, collected between 2020 2022, were analyzed. UAV multispectral images experimental area, vegetation indices extracted analyze correlation selected values. The input variables model screened, a support vector machine (SVM) was constructed estimate during heading, flowering, filling under water stresses. aim provide method rapid acquisition results showed that normal irrigation higher than those restriction. Multiple significantly correlated with prediction construction SPAD, models had high estimation both limitation treatments, coefficients predicted measured environments value r 0.59 0.81 RMSE 2.15 11.64, compared RE 0.10% 1.00%; drought stress environments, 0.69–0.79, 2.30–12.94, 0.10%–1.30%. This study demonstrated optimal combination feature selection methods learning algorithms can lead more accurate summary, SVM based rapidly accurately value.