文献类型： 外文期刊

作者： Fan Yi-guang ¹ ; Feng Hai-kuan ¹ ; Liu Yang ¹ ; Bian Ming-bo ¹ ; Zhao Yu ¹ ; Yang Gui-jun ¹ ; Qian Jian-guo ⁵ ;

作者机构： 1.Beijing Acad Agr & Forestry Sci, Informat Technol Res Ctr, Key Lab Quantitat Remote Sensing Agr, Minist Agr & Rural Affairs, Beijing 100097, Peoples R China

2.Nanjing Agr Univ, Natl Engn & Technol Ctr Informat Agr, Nanjing 210095, Peoples R China

3.Natl Engn Res Ctr Informat Technol Agr, Beijing 100097, Peoples R China

4.China Agr Univ, Key Lab Modern Precis Agr Syst Integrat Res, Minist Educ, Beijing 100083, Peoples R China

5.Liaoning Tech Univ, Sch Mapping & Geog Sci, Fuxing 123000, Peoples R China

关键词： Unmanned aerial vehicle; Potato; Plantnitrogen content; Vegetation indices; High frequency information

期刊名称：SPECTROSCOPY AND SPECTRAL ANALYSIS （ 2022影响因子：0.7； 五年影响因子：0.6 ）

ISSN： 1000-0593

年卷期： 2023 年 43 卷 5 期

页码：

收录情况： SCI

摘要： Plant nitrogen content (PNC) is essential for evaluating crop growth and nutritional status. Obtaining crop PNC information quickly and accurately can provide an important basis for formulating and implementing farmland management strategies. Existing studies have shown saturation in estimating crop PNC using only the spectral information of images. Therefore, this research attempted to use vegetation indices ( VIs) combined with two-dimensional discrete wavelet decomposition technology (DWT) to extract high-frequency information(HFI) at multiple scales. It was constructing a spectral, spatial feature (VIs-4 HFI) and exploring the ability of VIs, HFI, and VIs-4 HFI to estimate PNC. First, the UAV was a remote sensing platform to obtain digital images of the five critical nitrogennutrient growth periods of potato budding, tuber formation, tuber growth, starch accumulation, and maturity. It measured PNC data for each growth period. Secondly, based on the pre-processed UAV images, the spectral information of the canopy of each growth period was extracted to construct VIs, and the DWT was used to extract the HFI of each growth period 1-5 scales. Then, the VIs and HFI extracted from each growth period were correlated with the ground-truthed PNC data. The top 7 VIs and the top 10 HFI with larger absolute correlation coefficient values were screened, respectively. To reduce the effect of covariance on the experimental results, the screened HFI were subjected to principal component analysis (PCA) for dimensionality reduction according to the KMO test results. Finally, two methods, ridge regression and extreme learning machine (ELM), were used to construct and evaluate the PNC estimation model of each growth period of potato with VIs, HFI principal components, and VIs HFI principal components as model variables. The results showed that: (1) HFI at different scales contributed to the estimation of PNC in each growth period of potato. (2) The accuracy and stability of the potato PNC estimation model for each growth period constructed with VIs+HFI as model variables werehigher than that of a single VIs and HFI. (3) In each growth period of the potato, the PNC estimation model constructed by the ridge regression method was better than the ELM method. Among them, the PNC estimation model constructed with VIs+HFI as the model variable had the best effect. The modeling R-2 of the five growth periods were 0. 833, 0. 764, 0. 791, 0. 664, 0.435, and the RMSE were 0. 332% 0.297 %,0.275 %,0.286 %,0. 396%; NRMSE were 9. 113% 9. 425% 10. 336% 9. 547%, 15. 166% respectively. This research can provide new technical support for real-time and efficient potato nitrogen nutrition status detection.