欢迎访问《麦类作物学报》编辑部网站

作者	单位
岳蓉，尹本酥，李振发，李粉玲	（1.西北农林科技大学资源环境学院，陕西杨凌 712100 2.农业部西北植物营养与农业环境重点实验室，陕西杨凌 712100）

摘要点击次数: 73

全文下载次数: 103

中文摘要:

小麦植株感染条锈病后叶片花青素含量会发生明显变化。为了在地块尺度上利用冬小麦花青素值实现条锈病害的直观、快速监测，通过监测叶片花青素含量评估小麦条锈病严重程度，2021年获取感染条锈病的小麦田块的无人机RGB影像和采集地面病害区域的花青素含量数据，利用影像提取采样点感兴趣区的光谱特征参数和基于灰度共生矩阵的纹理特征参数，采用连续投影算法(SPA)结合相关性分析优选特征参数，分别采用单一光谱特征参数和组合参数，结合主成分回归(PCR)、拉索回归(LR)、随机森林回归(RFR)、梯度提升回归(GBR)和误差反向传播神经网络(BPNN)等方法构建了小麦花青素含量估算模型，并利用最优模型反演了田块的花青素含量。结果表明，图像光谱特征结合纹理特征后，花青素估算模型的R²增大，RMSE减小，模型精度显著提升。基于组合特征参数构建的随机森林模型精度最高，验证集R²、RMSE和MAE分别为0.801、0.026、0.021。该模型具有良好的花青素含量估算能力，得到的花青素值分布图与条锈病的空间分布具有一致性，能够定量化、可视化地反映病害严重程度。

英文摘要:

Yellow rust seriously affects the yield and quality of wheat, which causes changes not only in the external morphology and color characteristics of the plant, but also in its internal moisture, structure and pigments. Anthocyanin is one of the important pigments in the plant and would accumulate when the plant is under stress. Understanding and monitoring the anthocyanin of plants is of great importance for field management. In recent years, with the advantages of high resolution, high flexibility and low cost, UAV remote sensing technology is developing rapidly in the estimation of physical and chemical parameters of crops. In this study, taking anthocyanin content as the indicator to characterize the yellow rust, UAV RGB images of wheat plots infected by Puccinia striiformis f.sp.tritici in 2021 and anthocyanin content data were collected. The spectral features and texture features based on grey level co occurrence matrix of the regions of interest were extracted, and the correlation analysis between the features and anthocyanin content was carried out. The successive projections algorithm (SPA) combined with correlation analysis was employed to select the optimal features. From the 56 parameters, nine parameters were screened as the input variables of the models, including two texture features (MEA_R and MEA_G) and seven spectral features (red green ratio index, red blue ratio index, blue red ratio index, green blue ratio index, normalized difference red blue index, visible atmospherically resistant index, and anthocyanin reflectance index). Then anthocyanin content estimation models based on single spectral features and combined features were established with principal component regression (PCR), Lasso regression (LR), random forest regression (RFR), gradient boosting regression (GBR) and back propagation neural network (BPNN) algorithms respectively. The coefficient of determination, root mean square error (RMSE), and mean absolute error (MAE) were used to compare the accuracy of each model. Finally, the optimal model was used to invert the anthocyanin content of the plots. The results showed that the coefficient of determination of models increased, the root mean square error decreased and the accuracy of models improved after taking texture features into account. Among the anthocyanin content estimation models of winter wheat established through nine features, the model based on random forest algorithm had the highest accuracy. The validation results showed that the coefficient of determination of the model was 0.801; the root mean square error was 0.026, and the mean absolute error was 0.021. Overall, this model could be used for the remote sensing mapping of the winter wheat anthocyanin content at a regional scale. The final inversion result of anthocyanin content was consistent with the spatial distribution of yellow rust and could reflect the damage level in a quantitative and visual way. This technology can be used to rapidly measure the wheat leaf anthocyanin content and monitor the stripe rust.

查看全文查看/发表评论下载PDF阅读器

关闭