Research on predicting early Fusarium head blight with asymptomatic wheat grains by micro-near infrared spectrometer.

Fusarium head blight (FHB) is considered one of the most serious fungal diseases of wheat. Fusarium resulted in yield losses and contamination of harvested grains with mycotoxins. Therefore, diagnosing Fusarium head blight in early asymptomatic wheat is vital. To detect early FHB, a micro-near-infrared spectrometer was used to collect the spectrum of wheat grains, and FHB infection of wheat was detected by combining chemometrics in the 900-1700 nm near-infrared spectral region. First, the obtained spectra were analysed accordingly, and the pre-processed data were compared. The modelling analysis was then performed using the support vector machine (SVM), random forest (RF), extreme gradient descent (XGBoost), Autokeras, and Autogluon (with SVM) algorithms. The results showed that SG smoothing with standard normal variate (SG + SNV) was the best pre-treatment method. In addition, after SG + SNV was combined with the Autogluon (with SVM) model, the optimal classification results were obtained, with an accuracy of 73.33 % and an F1 value of 72.86 %. Autogluon (with SVM) could prevent overfitting and optimize generalization. Then, this manuscript discusses the performance of the Autogluon (with SVM) model with different stacking layers. The results show that one stacking layer can obtain a classification model with excellent performance. These results indicated that the near infrared spectrum (NIR) has the potential for early detection of Fusarium head blight with asymptomatic early statements.

A Novel Tran_NAS Method for the Identification of Fe- and Mg-Deficient Pear Leaves from N- and P-Deficient Pear Leaf Data.

Trace element deficiency diagnosis plays a critical role in pear cultivation. However, high-quality diagnostic models are challenging to investigate, making it difficult to collect samples. Therefore, this manuscript developed a novel transfer learning method, named Tran_NAS, with a fine-tuning neural network that uses a neural architecture search (NAS) to transfer learning from nitrogen (N) and phosphorus (P) to iron (Fe) and magnesium (Mg) to diagnose pear leaf element deficiencies. The best accuracy of the transferred NAS model is 89.12%, which is 11% more than that of the model without the transfer of trace element-deficient samples. Meanwhile, Tran_NAS also has better performance on source datasets after comparing with different proportions of training sets. Finally, this manuscript summarizes the transfer model coincident characteristics, including the methods of batch normalization (BN) and dropout layers, which make the model more generalizable. This manuscript applies a symmetric homogeneous feature-based transfer learning method on NAS that is designed explicitly for near-infrared (NIR) data collected from nutrient-deficient pear leaves. The novel transfer learning method would be more effective for the micro-NIR spectrum of the nondestructive diagnosis.