This study employed integrated multi-omics approaches to elucidate, from the perspective of amino acid metabolism, the adaptive mechanism of Penicillium digitatum under modified atmosphere packaging (MAP) conditions. Comparative analysis of natural air (Air), controlled atmosphere (CA), and MAP treatments revealed that MAP upregulated the expression of the hercynylcysteine S-oxide synthase (HCSOS), aldehyde dehydrogenase (ALDH), and monoamine oxidase (MAO) genes, thereby enhancing histidine-derived ergothioneine and methionine levels, and subsequently boosting glutathione-mediated redox homeostasis. Meanwhile, MAP induced the expression of the dihydroxyacid dehydratase (DHAD), saccharopine dehydrogenase (SDH), and arginosuccinate lyase (ASL) genes, redirecting valine, lysine, and arginine into the tricarboxylic acid (TCA) cycle to fuel ATP production. MAP also enhanced ASL-mediated arginine degradation and urea cycle activity, reducing arginine accumulation when compared to CA treatment. In contrast, while MAP induced upregulated expression of the pyrroline-5-carboxylate dehydrogenase (P5CDH) and D-amino acid oxidase (DAAO) genes, CA treatment promoted proline accumulation, reflecting stress-specific metabolic flexibility. Collectively, these findings demonstrate that MAP triggers transcriptional reprogramming of amino acid metabolism to coordinate oxidative defense, energy generation, and osmotic balance. By modulating these metabolic pathways and regulatory genes under MAP conditions, fungal adaptability can be disrupted. Hence, this study provides a promising strategy for suppressing green mold development, extending the postharvest shelf life, and improving the quality of fruits and vegetables.

Soil organic matter (SOM) is one of the essential components of soil and moisture to interfere with the detection. The soil moisture content by mass ratio and spectral parameters are often utilized to classify the moisture levels. But previous studies on the prediction of SOM content have focused mostly on the surface soil without covering the profile of depth range. This study aims to explore the influence of soil moisture on the inversion accuracy of the SOM profile. This study analyzed previously obtained near-infrared spectra were collected from the samples of the soil core column in the surface layer to about 150 cm underground. Each core was divided into subsamples with a height of 10 or 20 cm. The samples were gradually moistened to 6 groups of levels of soil moisture. According to the index of moisture tension, the air-dry state was defined as 1.500, 0.330, 0.100, 0.033 MPa, and the saturated state in turn. The data was normalized and transformed to the absorbance. Each set of spectral data was processed by seven spectral preprocessing. Among them, the standard normal variate transformation (SNV) was achieved the best. Meanwhile, successive projection algorithms (SPA) and competitive adaptive reweighting-successive projection algorithms (CARS-SPA) were used to screen the characteristic wavelengths. The number of characteristic wavelengths was then reduced to 6-8 at each level of soil moisture. The inversion models of SOM were constructed to combine with SNV preprocessing using full spectrum and characteristic wavelengths. The results indicated that: 1) The model accuracies of the SPA-PLSR and CARS-SPA-PLSR models were better than that of the PLSR model at six levels of moisture. 2) The SNV preprocessing was also combined to determine the coefficient of determination in prediction (R²_p) and root mean square error of the prediction set (RMSEP). Under the saturated state, R²_p values of the SNV-SPA-PLSR and SNV-CARS-SPA-PLSR models were 0.664 and 0.651, respectively, while the RMSEP values were 1.095 and 1.131 g/kg, respectively. In the air-dry state, R²_p values of the two models were 0.799 and 0.753, respectively, and RMSEP values were 0.759 and 0.848 g/kg, respectively, indicating the better prediction of the SNV-SPA-PLSR model. The SNV-CARS-SPA-PLSR model shared the higher accuracy of prediction when the moisture tension levels were 0.033, 0.100, 0.330, and 1.500 MPa. Specifically, R²_p values increased from 0.699 to 0.846, whereas, the RMSEP values decreased from 1.013 to 0.620 g/kg. 3) However, it was difficult to guarantee the same soil moisture level at the same depth in different locations of the field. The SOM calibration model was applied to the characteristic wavelengths on different datasets. The SNV-CARS-SPA-PLSR model was selected at the moisture tension of 1.500 MPa. The best performance was achieved for the organic matter in the six groups of soil moisture levels and mixed samples. The models can be expected to estimate the SOM content of the profile at various moisture levels. The findings can also provide a strong reference to improve the applicability of near-infrared spectra inversion models for the organic matter content at different levels of soil moisture.

As the negative impact of weeds on agricultural production becomes increasingly severe, efficient weed detection methods are crucial for promoting sustainable agricultural development. However, existing weed detection models still face challenges in terms of accuracy and efficiency. To address this issue, this paper, based on YOLOv5, introduced Knowledge Distillation techniques and integrated the "knowledge" from multiple-teacher models, aiming to enhance performance and real-time detection ability of the weed detection model. Firstly, a temperature coefficient-based soft voting mechanism for logits (TS_Logits) was proposed, which dynamically adjusted the distillation loss weights of the teacher models, effectively integrating the strengths of each teacher model. Meanwhile, a multi-teacher feature fusion method based on attention mechanisms (AT_Feature) was also proposed, which dynamically weighted key features and suppressed redundant information. In this paper, the open-source CottonWeedDet12 dataset was used, with data augmentation techniques applied for random expansion, ultimately obtaining 9210 weed images. These images were divided into training and validation sets in an 8:2 ratio, and a test set consisting of 554 actual weed images was also incorporated. Based on YOLOv5s, channel pruning techniques were applied to obtain a student model with a size of 2.9 MB, an F1-score (F₁) of 94.5%, and a mean average precision of validation set (mAP_val) of 96.8%. Meanwhile, YOLOv5s, YOLOv7 and YOLOv10s were used as teacher models, and three Logits-based knowledge distillation methods (KD, Luminet, and DKD) and ten Feature-based knowledge distillation methods (FitNet, AT, NST, PKT, RKD, VID, SemCKD, CWD, MGD, and FGD) were applied. The experimental results showed that the student model achieved the best performance when YOLOv5s was used as the teacher model, further proving that using the same model architecture facilitated knowledge transfer between the teacher and student models, thereby enhancing the distillation effect. Additionally, this study evaluated the effect of the TS_Logits and the AT_Feature on the student model performance. The TS_Logits, which combined the strengths of YOLOv5s and YOLOv7, significantly enhanced the student model performance, achieving an F₁ of 94.2% and a mAP_val of 96.4%. The AT_Feature significantly enhanced the student model performance by integrating features from the teacher models, generating richer feature maps. Specifically, MGD achieved an F₁ of 93.9% and a mAP of 96.4%; CWD achieved an F₁ of 94.2% and a mAP_val of 96.4%; and PKT achieved an F₁ of 94.3% and a mAP_val of 96.4%. After combining the TS_Logits and AT_Feature, the proposed YOLOv5s-MGD performed excellently in model accuracy and real-time detection performance. The model achieved an F₁ of 94.5%, a mAP_val of 96.8%, a mean average precision of test set (mAP_test) of 93.6%, a frame rate of 46.71 frames per second (FPS), a computational complexity of 4.1 giga floating point operations (GFLOPs), and a model size of 2.9 MB. Compared to YOLOv5s, YOLOv5s-MGD exhibited a decrease of only 0.9 percentage points in mAP_val, while the FPS increased by approximately 57.22%, computational complexity was reduced by about 74.38%, and the model size decreased by approximately 79.86%. Compared to YOLOv7, YOLOv5s-MGD showed a 1.3 percentage points decrease in mAP_val, but FPS increased by approximately 1076.57%, along with significant optimizations in computational complexity and model size. Compared to YOLOv10s, YOLOv5s-MGD had a 0.9 percentage points decrease in mAP_val, FPS increased by approximately 143.41%, computational complexity was reduced by about 83.27%, and the model size decreased by approximately 82.42%. In summary, YOLOv5s-MGD not only maintained high accuracy but also significantly improved detection speed, computational efficiency, and storage performance, making it well-suited for real-time deployment on resource-constrained devices. This model effectively addressed the challenges of weed detection and provided strong technical support for the development of modern agriculture.