Multimodal deep learning-based drought monitoring research for winter wheat during critical growth stages.

Wheat is a major grain crop in China, accounting for one-fifth of the national grain production. Drought stress severely affects the normal growth and development of wheat, leading to total crop failure, reduced yields, and quality. To address the lag and limitations inherent in traditional drought monitoring methods, this paper proposes a multimodal deep learning-based drought stress monitoring S-DNet model for winter wheat during its critical growth periods. Drought stress images of winter wheat during the Rise-Jointing, Heading-Flowering and Flowering-Maturity stages were acquired to establish a dataset corresponding to soil moisture monitoring data. The DenseNet-121 model was selected as the base network to extract drought features. Combining the drought phenotypic characteristics of wheat in the field with meteorological factors and IoT technology, the study integrated the meteorological drought index SPEI, based on WSN sensors, and deep image learning data to build a multimodal deep learning-based S-DNet model for monitoring drought stress in winter wheat. The results show that, compared to the single-modal DenseNet-121 model, the multimodal S-DNet model has higher robustness and generalization capability, with an average drought recognition accuracy reaching 96.4%. This effectively achieves non-destructive, accurate, and rapid monitoring of drought stress in winter wheat.

Identification of winter wheat pests and diseases based on improved convolutional neural network.

Wheat pests and diseases are one of the main factors affecting wheat yield. According to the characteristics of four common pests and diseases, an identification method based on improved convolution neural network is proposed. VGGNet16 is selected as the basic network model, but the problem of small dataset size is common in specific fields such as smart agriculture, which limits the research and application of artificial intelligence methods based on deep learning technology in the field. Data expansion and transfer learning technology are introduced to improve the training mode, and then attention mechanism is introduced for further improvement. The experimental results show that the transfer learning scheme of fine-tuning source model is better than that of freezing source model, and the VGGNet16 based on fine-tuning all layers has the best recognition effect, with an accuracy of 96.02%. The CBAM-VGGNet16 and NLCBAM-VGGNet16 models are designed and implemented. The experimental results show that the recognition accuracy of the test set of CBAM-VGGNet16 and NLCBAM-VGGNet16 is higher than that of VGGNet16. The recognition accuracy of CBAM-VGGNet16 and NLCBAM-VGGNet16 is 96.60 and 97.57%, respectively, achieving high precision recognition of common pests and diseases of winter wheat.

Nanocomposite of Half-Fin Anchovy Hydrolysates/Zinc Oxide Nanoparticles Exhibits Actual Non-Toxicity and Regulates Intestinal Microbiota, Short-Chain Fatty Acids Production and Oxidative Status in Mice.

The nanocomposite of half-fin anchovy hydrolysates (HAHp) and zinc oxide nanoparticles (ZnO NPs) (named as HAHp(3.0)/ZnO NPs) demonstrated increased antibacterial activity compared to either HAHp(3.0) or ZnO NPs as per our previous studies. Also, reactive oxygen species (ROS) formation was detected in Escherichia coli cells after treatment with HAHp(3.0)/ZnO NPs. The aim of the present study was to evaluate the acute toxicity of this nanocomposite and to investigate its effect on intestinal microbiota composition, short-chain fatty acids (SCFAs) production, and oxidative status in healthy mice. The limit test studies show that this nanoparticle is non-toxic at the doses tested. The administration of HAHp(3.0)/ZnO NPs, daily dose of 1.0 g/kg body weight for 14 days, increased the number of goblet cells in jejunum. High-throughput 16S ribosomal RNA gene sequencing of fecal samples revealed that HAHp(3.0)/ZnO NPs increased Firmicutes and reduced Bacteriodetes abundances in female mice. Furthermore, the microbiota for probiotic-type bacteria, including Lactobacillus and Bifidobacterium, and SCFAs-producing bacteria in the Clostridia class, e.g., Lachnospiraceae_unclassified and Lachnospiraceae_UCG-001, were enriched in the feces of female mice. Increases of SCFAs, especially statistically increased propionic and butyric acids, indicated the up-regulated anti-inflammatory activity of HAHp(3.0)/ZnO NPs. Additionally, some positive responses in liver, like markedly increased glutathione and decreased malonaldehyde contents, indicated the improved oxidative status. Therefore, our results suggest that HAHp(3.0)/ZnO NPs could have potential applications as a safe regulator of intestinal microbiota or also can be used as an antioxidant used in food products.

[Artificial neural networks for the identification of infrared spectra of Ilex Kudingcha].

In order to identify Ilex Kudingcha, two kinds of models of artificial neural networks (ANN), i.e. competitive neural network and back propagation neural network, were used to analyze their infrared spectra. Ilex Kudingcha samples were collected by Fourier transform infrared (FTIR) spectra. Twenty five samples were gathered as a train set, and 11 samples as a test set, then their training was performed using two networks each. The results show that the identification of Ilex Kudingcha from different areas can be effectively performed with the competitive neural network and BP network, but the competitive neural network is used in the identification of different grades of Ilex Kudingcha. The results were better in training speed and accuracy with the competitive neural network. In conclusion, the competitive neural network combined with FTIR spectroscopy is a good method for the identification of Ilex Kudingcha.