Zifei Ma

Citrus fruit size is a core quality attribute that determines commercial value and market acceptance. In postharvest processing, rapid, accurate, and non-destructive diameter estimation is an essential prerequisite for automated grading, yet existing computer vision methods are difficult to deploy on factory edge devices. To address this problem, this paper proposes PMDS-YOLO, a lightweight deep learning model for non-destructive citrus fruit diameter estimation and automated quality grading. The model incorporates four key improvements: width factor compression to 0.125 to eliminate channel redundancy; adoption of a MobileOne backbone embedded with the proposed multi-scale channel–spatial residual attention module (MCSR), which reduces parameter count while enhancing multi-scale feature extraction and background discrimination; introduction of dynamic snake convolution (DySnakeConv) to improve the feature extraction module for better detection of irregular fruit contours; and adoption of soft non-maximum suppression (Soft-NMS) to reduce missed detections of densely overlapping targets. Experimental results on a self-built citrus dataset show that PMDS-YOLO achieves only 0.58 M parameters and 86.41% mAP@0.5:0.95, representing a 76.4% reduction in parameter count and a 0.99 percentage point improvement in mAP over the baseline YOLOv13n. In the diameter estimation stage, a quadratic polynomial regression calibration model reduces the MAE to 2.15 mm (R² = 0.85, RMSE = 3.16 mm) and achieves an overall grading accuracy of 57.0% in the grading evaluation, indicating that while the method demonstrates potential in diameter estimation accuracy, further optimization is required for grading performance. The proposed system provides a low-cost, edge-deployable solution for preliminary fruit size assessment and small-scale sorting applications.

Monitoring water surface algal blooms from surveillance perspectives faces challenges such as small objects, low texture contrasts, dynamic background interferences, and limited labeled datasets. In this study, we propose GECA-YOLOv26, a lightweight model that integrates Ghost Convolution (GhostConv) and Efficient Channel Attention (ECA) modules. First, the GhostConv lightweight module is introduced in the first layer of the YOLOv26 backbone, reducing parameters from 4608 to 2704 and achieving a 41% reduction in computational cost. Second, eight ECA modules are embedded at key locations after backbone downsampling and neck feature fusion to enhance feature representation and mitigate degradation caused by model lightweighting. Finally, the MuSGD optimizer is used for training, with adaptive modifications to resolve tensor shape conflicts with the ECA modules. Experimental results indicate that the model achieves a mAP50 of 82.16%. Compared with the YOLOv26 baseline, our model improves mAP50 by 6.42%, while mAP@0.5:0.95 decreases by 0.79% and inference speed reduces from 143 FPS to 123 FPS. The model also reduces parameters and size, achieving 5.19 MB and 1864 fewer parameters. Compared with YOLOv8, YOLOv10, and YOLOv11, the proposed model improves mAP50 by 2.12%, 5.99%, and 2.79%, respectively. To evaluate the stability of the results under small-sample conditions, we conducted 3-fold and 5-fold cross-validation experiments, which demonstrated that the model performs robustly across different folds and random seeds. Ablation studies further confirm the effectiveness of each module. Heatmap analysis demonstrates that the proposed model effectively highlights small object regions, remains robust under limited-sample conditions, and reduces model complexity. This study provides a novel solution for algal bloom detection in surveillance scenarios.

P-wave first-motion polarity is a critical parameter for determining earthquake focal mechanisms. Extracting relative P-wave arrival times and polarity information using waveform cross-correlation techniques can enhance the accuracy of earthquake location and focal mechanism inversion. However, seismic noise severely hampers the reliable detection of P-wave onsets and their first-motion polarities. To address this issue, we propose a noise-resistant polarity detection method based on the normal time–frequency transform (NTFT), termed the time–frequency similarity coefficient (TFSC). The TFSC method computes relative delays and similarity coefficients by calculating the real part of the NTFT coefficients between two seismic signals. We validated the proposed approach using both synthetic and real earthquake data. Without any filtering or preprocessing, the TFSC method demonstrated significantly greater robustness and reliability compared to the conventional time-domain normalized cross-correlation (NCC) method. These results indicate that the TFSC method has strong potential for practical application and provides a novel perspective for accurate detection of P-wave first-motion polarity in noisy seismic environments.

Urban–Rural Water supply integration is one of the effective ways to address rural drinking water safety issues. With the rapid economic development and urbanization, the gap between urban and rural water supply has become increasingly evident, and rural areas face severe challenges such as insufficient water sources and substandard water quality. Yunnan Province, due to its unique topography and natural environment, encounters numerous difficulties in promoting the construction and development of urban–rural water supply integration, and there is a severe lack of research specifically focused on sustainable development measures for urban–rural water supply integration in Yunnan Province. In light of this, this paper first reviews the current research status on urban–rural water supply integration both domestically and internationally, pointing out that existing studies mainly focus on optimizing urban water supply systems, addressing rural drinking water safety issues, and exploring water supply management models. It then analyzes the challenges faced by Yunnan Province in advancing urban–rural water supply integration, including limitations in engineering construction, inadequate operation and maintenance, complex and variable water source conditions, and insufficient public awareness of water conservation. In response to these issues, a series of sustainable development measures are proposed, including revitalizing existing assets, enhancing construction planning and quality, promoting water pricing reforms, upgrading the technical system for drinking water safety assurance, and advancing the information technology construction of urban–rural water supply integration, with the aim of providing references and insights for the sustainable development of urban–rural water supply integration in Yunnan Province and other regions.

A digital twin is a new trend in the development of the current smart water conservancy industry. The main research content of intelligent water conservancy is clarified. This paper first summarizes and combs the relevant system architecture of smart water conservancy, and puts forward a smart water conservancy framework based on digital twins, highlighting the characteristics of virtual and real interaction, and symbiosis of the water conservancy twin platform. Secondly, the status quo of intelligent water conservancy “sky, air, ground and water” integrated monitoring technology, big data and artificial intelligence, model platform technology, knowledge graph and security technology is analyzed. From the perspective of application, the research progress of each technology in water security, water resources and hydraulic engineering is reviewed. Although the construction of smart water conservancy has made remarkable progress, it still faces many challenges such as data governance, technology integration and innovation, and standardization. In view of these challenges, this paper puts forward a series of countermeasures, and looks forward to the future development direction of intelligent water conservancy.

Digital twin technology, a new type of digital technology emerging in recent years, realizes real-time simulation, prediction and optimization by digitally modeling the physical world, providing a new idea and method for the design, operation and management of water conservancy projects, which is of great significance for the realization of the transformation of water conservancy informatization to intelligent water conservancy. In view of this, this paper systematically discusses the concept and development history of digital twin smart water conservancy, compares its differences with traditional water conservancy models, and further proposes the digital twin smart water conservancy five-dimensional model. Based on the five-dimensional model of digital twin water conservancy, the research progress of digital twin smart water conservancy is summarized by focusing on six aspects, namely digital twin water conservancy data perception, data transmission, data analysis and processing, digital twin water conservancy model construction, digital twin water conservancy interaction and collaboration and digital twin water conservancy service application, and the challenges and problems of digital twin technology in the application of smart water conservancy. Finally, the development trend of digital twin technology and the direction of technological breakthroughs are envisioned, aiming to provide reference and guidance for the research on digital twin technology in the field of smart water conservancy and to promote the further development of the field.

Crop diseases pose a major threat to agricultural production, quality, and sustainable development, highlighting the importance of early disease risk prediction for effective disease control. Tea anthracnose can easily occur in Yunnan under high-temperature and high-humidity environments, which seriously affects the ecosystem of tea gardens. Therefore, the establishment of accurate, non-destructive, and rapid prediction models has a positive impact on the conservation of biodiversity in tea plantations. Because of the linear relationship between disease occurrence and environmental conditions, the growing environmental conditions can be effectively used to predict crop diseases. Based on the climate data collected by Internet of Things devices, this study uses LASSO-COX-NOMOGRAM to analyze the expression of tea anthracrum to different degrees through Limma difference analysis, and it combines Cox single-factor analysis to study the influence mechanism of climate and environmental change on tea anthracrum. Modeling factors were screened by LASSO regression, 10-fold cross-validation and Cox multi-factor analysis were used to establish the basis of the model, the nomogram prediction model was constructed, and a Shiny- and DynNOM-visualized prediction system was built. The experimental results showed that the AUC values of the model were 0.745 and 0.731 in the training set and 0.75 and 0.747 in the verification set, respectively, when the predicted change in tea anthracnose disease risk was greater than 30% and 60%, and the calibration curve was in good agreement with the ideal curve. The accuracy of external verification was 83.3% for predicting tea anthracnose of different degrees. At the same time, compared with the traditional prediction method, the method is not affected by the difference in leaf background, which provides research potential for early prevention and phenotypic analysis, and also provides an effective means for tea disease identification and harm analysis.

Deep learning models are widely used in crop leaf disease image recognition. These models can be divided into two categories: global model and local model. The global model directly takes the whole leaf disease images as input to training and recognition. It can achieve end-to-end training and recognition, which is very convenient to use. But this kind of model cannot very accurately and completely extract the features from the very small diseased spots in the image. Before training and recognizing, the local model needs to extract the diseased spots part from the image by image segmentation technology. Then the local model takes the disease spots part images as input to training and recognition. Features extracted by local model are more accurate and complete. But this kind of model cannot achieve end-to-end training and recognition, and the image segmentation will bring additional overhead. Considering the disadvantage of global model and local model, we proposed a crop leaf disease image recognition method based on bilinear residual networks (named DIR-BiRN). DIR-BiRN extracts features by two residual networks feature extractors and then integrates the features by a bilinear pooling function. By this way, it can extract features more accurately and completely while achieving end-to-end training and recognition. Experiments on the PlantVillage dataset show that, when compared with the standard ResNet-18 model, the DIR-BiRN improves on accuracy performance, recall performance, precision performance, and F1-measure performance by averages of 0.2918, 0.81641, 0.59185, and 0.52151 percentage points, respectively.

Software knowledge community contains a large scale of software knowledge entities with complex structure and rich semantic relations. Semantic relation extraction of software knowledge entities is a critical task for software knowledge graph construction, which has an important impact on knowledge graph based tasks such as software document generation and software expert recommendation. Due to the problems of entity sparsity, relation ambiguity, and the lack of annotated dataset in user-generated content of software knowledge community, it is difficult to apply existing methods of relation extraction in the software knowledge domain. To address these issues, we propose a novel software knowledge entity relation extraction model which incorporates entity-aware information with syntactic dependency information. Bidirectional Gated Recurrent Unit (Bi-GRU) and Graph Convolutional Networks (GCN) are used to learn the features of contextual semantic representation and syntactic dependency representation, respectively. To obtain more syntactic dependency information, a weight graph convolutional network based on Newtons cooling law is constructed by calculating a weight adjacency matrix. Specifically, an entity-aware attention mechanism is proposed to integrate the entity information and syntactic dependency information to improve the prediction performance of the model. Experiments are conducted on a dataset which is constructed based on texts of the StackOverflow and show that the proposed model has better performance than the benchmark models.