Sugarcane Nitrogen and Irrigation Level Prediction Based on UAV-Captured Multispectral Image at Elongating Stage

Abstract

Abstract Introduction Sugarcane is the main industrial crop for sugar production; its growth status is closely related to fertilizer, water, and light input. Unmanned aerial vehicle (UAV)-based multispectral imagery is widely used for high-throughput phenotyping because it can rapidly predict crop vigor. This paper mainly studied the potential of multispectral images obtained by low-altitude UAV systems in predicting canopy nitrogen (N) content and irrigation level for sugarcane. Methods An experiment was carried out on sugarcane fields with three irrigation levels and five nitrogen levels. A multispectral image at a height of 40 m was acquired during the elongation stage, and the canopy nitrogen content was determined as the ground truth. N prediction models, including partial least square (PLS), backpropagation neural network (BPNN), and extreme learning machine (ELM) models, were established based on different variables. A support vector machine (SVM) model was used to recognize the irrigation level. Results The PLS model based on band reflectance and five vegetation indices had better accuracy (R=0.7693, root mean square error (RMSE)=0.1109) than the BPNN and ELM models. Some spectral information from the multispectral image had obviously different features among the different irrigation levels, and the SVM algorithm was used for irrigation level classification. The classification accuracy reached 77.8%. Conclusion Low-altitude multispectral images could provide effective information for N prediction and water irrigation level recognition.

Plant phenotyping relevance

UAVマルチスペクトル画像からサトウキビのキャノピー窒素含量と灌漑レベルを推定・分類する手法が研究の中心であり、モデル精度も評価しているため。

Code and data availability

The supplied blocks describe UAV multispectral image acquisition, N prediction models (PLS/BPNN/ELM), and SVM irrigation classification, but contain no data availability statement, no public dataset or image deposit, and no author code/model release. No paper-specific public asset is identified.

No evidence-backed public reproduction asset is currently recorded.