Yield estimation using remote sensing data is a research priority in modern agriculture. The rapid and accurate estimation of winter wheat yields over large areas is an important prerequisite for food security policy formulation and implementation. In most county-level yield estimation processes, multiple input data are used for yield prediction as much as possible, however, in some regions, data are more difficult to obtain, so we used the single-leaf area index (LAI) as input data for the model for yield prediction. In this study, the effects of different time steps as well as the LAI time series on the estimation results were analyzed for the properties of long short-term memory (LSTM), and multiple machine learning methods were compared with yield estimation models constructed by the LSTM networks. The results show that the accuracy of the yield estimation results using LSTM did not show an increasing trend with the increasing step size and data volume, while the yield estimation results of the LSTM were generally better than those of conventional machine learning methods, with the best R² and RMSE results of 0.87 and 522.3 kg/ha, respectively, in the comparison between predicted and actual yields. Although the use of LAI as a single input factor may cause yield uncertainty in some extreme years, it is a reliable and promising method for improving the yield estimation, which has important implications for crop yield forecasting, agricultural disaster monitoring, food trade policy, and food security early warning.

Aiming at the problem of accurately locating and identifying multi-scale and differently shaped apple leaf diseases from a complex background in natural scenes, this study proposed an apple leaf disease detection method based on an improved YOLOv5s model. Firstly, the model utilized the bidirectional feature pyramid network (BiFPN) to achieve multi-scale feature fusion efficiently. Then, the transformer and convolutional block attention module (CBAM) attention mechanisms were added to reduce the interference from invalid background information, improving disease characteristics’ expression ability and increasing the accuracy and recall of the model. Experimental results showed that the proposed BTC-YOLOv5s model (with a model size of 15.8M) can effectively detect four types of apple leaf diseases in natural scenes, with 84.3% mean average precision (mAP). With an octa-core CPU, the model could process 8.7 leaf images per second on average. Compared with classic detection models of SSD, Faster R-CNN, YOLOv4-tiny, and YOLOx, the mAP of the proposed model was increased by 12.74%, 48.84%, 24.44%, and 4.2%, respectively, and offered higher detection accuracy and faster detection speed. Furthermore, the proposed model demonstrated strong robustness and mAP exceeding 80% under strong noise conditions, such as exposure to bright lights, dim lights, and fuzzy images. In conclusion, the new BTC-YOLOv5s was found to be lightweight, accurate, and efficient, making it suitable for application on mobile devices. The proposed method could provide technical support for early intervention and treatment of apple leaf diseases.

Utilizing image data for yield estimation is a key topic in modern agriculture. This paper addresses the difficulty of counting wheat spikelets using images, to improve yield estimation in wheat fields. A wheat spikelet image dataset was constructed with images obtained by a smartphone, including wheat ears in the flowering, filling, and mature stages of reproduction. Furthermore, a modified lightweight object detection method, YOLOv5s-T, was incorporated. The experimental results show that the coefficient of determination ($R^2$) between the predicted and true values of wheat spikelets was 0.97 for the flowering stage, 0.85 for the grain filling stage, and 0.78 for the mature stage. The $R^2$ in all three fertility stages was 0.87, and the root mean square error (RMSE) was 0.70. Compared with the original YOLOv5s algorithm, the spikelet detection counting effect of YOLOv5s-T was not reduced. Meanwhile, the model size was reduced by 36.8% (only 9.1 M), the GPU memory usage during the training process was reduced by 0.82 GB, the inference time was reduced by 2.3 ms, the processing time was reduced by 10 ms, and the calculation amount was also reduced. The proposed YOLOv5s-T algorithm significantly reduces the model size and hardware resource requirements while guaranteeing high detection and counting accuracy, which indicates the potential for wheat spikelet counting in highly responsive wheat yield estimation.

Corn is one of the global staple grains and the largest grain crop in China. During harvesting, grain separation, and corn production, corn is susceptible to mechanical damage including surface cracks, internal cracks, and breakage. However, the internal cracks are difficult to observe. In this study, hyperspectral imaging was used to detect mechanical damage in corn seeds. The corn seeds were divided into four categories that included intact, broken, internally cracked, and surface-crackedtv. This study compared three feature extraction methods, including principal component analysis (PCA), kernel PCA (KPCA), and factor analysis (FA), as well as a joint feature extraction method consisting of a combination of these methods. The dimensionality reduction results of the three methods (FA + KPCA, KPCA + FA, and PCA + FA) were combined to form a new combined dataset and improve the classification. We then compared the effects of six classification models (ResNet, ShuffleNet-V2, MobileNet-V3, ResNeSt, EfficientNet-V2, and MobileNet-V4) and proposed a ResNeSt_E network based on the ResNeSt and efficient multi-scale attention modules. The accuracy of ResNeSt_E reached 99.0%, and this was 0.4% higher than that of EfficientNet-V2 and 0.7% higher than that of ResNeSt. Additionally, the number of parameters and memory requirements were reduced and the frames per second were improved. We compared two dimensionality reduction methods: KPCA + FA and PCA + FA. The classification accuracies of the two methods were the same; however, PCA + FA was much more efficient than KPCA + FA and was more suitable for practical detection. The ResNeSt_E network could detect both internal and surface cracks in corn seeds, making it suitable for mobile terminal applications. The results demonstrated that detecting mechanical damage in corn seeds using hyperspectral images was possible. This study provides a reference for mechanical damage detection methods for corn.

In the food domain, text mining techniques are extensively employed to derive valuable insights from large volumes of text data, facilitating applications such as aiding food recalls, offering personalized recipes, and reinforcing food safety regulation. To provide researchers and practitioners with a comprehensive understanding of the latest technology and application scenarios of text mining in the food domain, the pertinent literature is reviewed and analyzed. Initially, the fundamental concepts, principles, and primary tasks of text mining, encompassing text categorization, sentiment analysis, and entity recognition, are elucidated. Subsequently, an analysis of diverse types of data sources within the food domain and the characteristics of text data mining is conducted, spanning social media, reviews, recipe websites, and food safety reports. Furthermore, the applications of text mining in the food domain are scrutinized from the perspective of various scenarios, including leveraging consumer food reviews and feedback to enhance product quality, providing personalized recipe recommendations based on user preferences and dietary requirements, and employing text mining for food safety and fraud monitoring. Lastly, the opportunities and challenges associated with the adoption of text mining techniques in the food domain are summarized and evaluated. In conclusion, text mining holds considerable potential for application in the food domain, thereby propelling the advancement of the food industry and upholding food safety standards.

Fusarium head blight (FHB), a severe ailment triggered by fungal pathogens, poses a considerable risk to both the yield and quality of winter wheat worldwide, underscoring the urgency for precise detection measures that can effectively mitigate and manage the spread of FHB. Addressing the limitations of current deep learning models in capturing detailed features from UAV imagery, this study proposes an advanced identification model for FHB in wheat based on multispectral imagery from UAVs. The model leverages the U²Net network as its baseline, incorporating the Coordinate Attention (CA) mechanism and the RFB-S (Receptive Field Block—Small) multi-scale feature extraction module. By integrating key spectral features from multispectral bands (SBs) and vegetation indices (VIs), the model enhances feature extraction capabilities and spatial information awareness. The CA mechanism is used to improve the model’s ability to express image features, while the RFB-S module increases the receptive field of convolutional layers, enhancing multi-scale spatial feature modeling. The results demonstrate that the improved U²Net model, termed U²Net-plus, achieves an identification accuracy of 91.73% for FHB in large-scale wheat fields, significantly outperforming the original model and other mainstream semantic segmentation models such as U-Net, SegNet, and DeepLabV3+. This method facilitates the rapid identification of large-scale FHB outbreaks in wheat, providing an effective approach for large-field wheat disease detection.