Deep convolution neural network (DCNN) technology has achieved great success in extracting buildings from aerial images. However, the current mainstream algorithms are not satisfactory in feature extraction and classification of homesteads, especially in complex rural scenarios. This study proposes a deep convolutional neural network for rural homestead extraction consisting of a detail branch, a semantic branch, and a boundary branch, namely Multi-Branch Network (MBNet). Meanwhile, a multi-task joint loss function is designed to constrain the consistency of bounds and masks with their respective labels. Specifically, MBNet guarantees the details of prediction through serial 4× down-sampled high-resolution feature maps and adds a mixed-scale spatial attention module at the tail of the semantic branch to obtain multi-scale affinity features. At the same time, the low-resolution semantic feature maps and interaction between high-resolution detail feature maps are maintained. Finally, the result of semantic segmentation is refined by the point-to-point module (PTPM) through the generated boundary. Experiments on UAV high-resolution imagery in rural areas show that our method achieves better performance than other state-of-the-art models, which helps to refine the extraction of rural homesteads. This study demonstrates that MBNet is a potential candidate for building an automatic rural homestead management system.

In ridge-cultivated strawberry environments, significant scale variation between fruits and stems, severe occlusion, and high missed-detection rates for small or slender targets pose major challenges to accurate picking point localization. This study proposes a fruit-stem oriented detection and picking point localization method based on an improved YOLOv11n-OBB model. Specifically, an enhanced model termed FSGE-OBB is developed. A Feature Fusion Lite Module (FFLM) is introduced to strengthen cross-layer interaction between shallow spatial details and deep semantic features, thereby improving multi-scale feature representation. A Stem Direction Enhancement Module (SDEM) is designed to enhance the directional perception of slender stems. In addition, Grouped Spatial Excitation Convolution (GSEConv) is adopted to reduce model complexity while maintaining feature representation capability, and an Efficient Upsampling Convolution Block (EUCB) is incorporated into high-resolution feature maps to improve detail recovery. Based on the oriented detection results, a geometry-constrained fruit-stem association method is established by modeling the spatial relationship between fruit and stem oriented bounding boxes (OBBs), and the corresponding picking points are localized accordingly. Experimental results show that the FSGE-OBB model achieves mAP@0.5 and mAP@0.5:0.95 of 78.85% and 68.27%, respectively, outperforming the YOLOv11n-OBB baseline by 2.59 and 1.90 percentage points. The proposed fruit-stem association method achieves a success rate of 97.99%. In preliminary three dimensional localization experiments, the mean absolute errors of picking points are 2.2 mm, 2.6 mm, and 2.3 mm in the x- , y- , and z- directions, respectively. Furthermore, simulated harvesting experiments achieve a success rate of 93.3%, with an average harvesting time of 9.2 s per fruit, validating the feasibility and effectiveness of the proposed method in practical harvesting tasks. These results demonstrate that the proposed FSGE-OBB-based framework enables unified modeling of fruit-stem detection, association, and picking point localization, providing a feasible visual perception solution for strawberry harvesting robots.

Agricultural science and technology policies (ASTPs) have played a pivotal role in shaping agricultural innovation, sustainability, and cleaner production practices. Understanding how ASTPs diffuse is essential for optimizing policy design and advancing the green transition in agriculture. This study aims to investigate the diffusion of ASTPs in China, using a quantitative citation-based approach. The goal is to explore diffusion patterns, topic characteristics, and historical trajectories of ASTPs, thereby providing insights into policy transmission mechanisms that can inform future policy improvements. We analyze 3207 ASTP documents, focusing on policy citation links to examine the distribution, diffusion characteristics, and dynamics of policies. The analysis includes tracking topic evolution and identifying key policies while estimating the main diffusion paths. The results show that the top-down diffusion model is the dominant pattern of policy transmission, exhibiting the highest diffusion speed and both short- and long-term impacts. ASTPs have progressively expanded toward industrialization, informatization, and green development, with increased policy transmission efficiency. The diffusion process has formed three primary pathways: (i) enhancing agricultural innovation capacity, (ii) accelerating the transformation of technological achievements, and (iii) improving the agricultural science and technology innovation system. These pathways are critical to advancing sustainable and cleaner agricultural production. This study provides valuable insights into the diffusion of ASTPs and highlights key pathways for policy optimization. The findings suggest that enhancing policy frameworks and improving policy implementation efficiency will be crucial for facilitating the transition toward sustainable, low-carbon, and environmentally friendly agricultural practices. Future research should refine data sources and incorporate semantic analysis to capture more detailed policy transmission mechanisms.

The convergence of agricultural biotechnology and artificial intelligence is reshaping modern crop improvement. Despite a surge of studies integrating artificial intelligence and biotechnology, the rapidly expanding literature on intelligent crop breeding remains fragmented across molecular, phenotypic, and computational dimensions. Existing reviews often rely on traditional bibliometric or narrative approaches that fail to capture the deep semantic evolution of research themes. To address this gap, this study employs the BERTopic model to systematically analyze 1867 articles (1995–2025, WoS Core Collection), mapping the thematic landscape and temporal evolution of intelligent crop breeding and revealing how methodological and application-oriented domains have co-evolved over time. Eight core topics emerge, i.e., (T0) genomic prediction and genotype–environment modeling; (T1) UAV remote sensing and multimodal phenotyping; (T2) stress-tolerant breeding and root phenotypes; (T3) ear/pod counting with deep learning; (T4) grain trait representation and evaluation; (T5) CRISPR and genome editing; (T6) spike structure recognition and 3D modeling; and (T7) maize tassel detection and developmental staging. Topic-evolution analyses indicate a co-development pattern, where genomic prediction provides a stable methodological backbone, while phenomics (UAV/multimodal imaging, organ-level detection, and 3D reconstruction) propels application-oriented advances. Attention dynamics reveal increasing momentum in image-based counting (T3), grain quality traits (T4), and CRISPR-enabled editing (T5), alongside a plateau in traditional mainstays (T0, T1) and mild cooling in root phenotyping under abiotic stress (T2). Quality stratification (citation quartiles, Q1–Q4) shows high-impact concentration in T0/T1 and a growing tail of application-driven work across T3–T7. Journal analysis reveals a complementary publication ecosystem: Frontiers in Plant Science and Plant Methods anchor cross-disciplinary dissemination; Remote Sensing and Computers and Electronics in Agriculture host engineering-centric phenomics; genetics/breeding journals sustain T0/T2; and molecular journals curate T5. These findings provide an integrated overview of methods, applications, and publication venues, offering practical guidance for research planning, cross-field collaboration, and translational innovation in intelligent crop breeding.

Accurate maize yield prediction across diverse environments is pivotal for modern breeding programs. While machine learning (ML) excels at capturing non-linear environmental effects, Genomic Best Linear Unbiased Prediction (GBLUP) remains a benchmark for modeling polygenic small-effect contributions. However, principled integration of these paradigms—while explicitly accounting for genotype-by-environment interaction (G×E)—remains a formidable challenge. We propose a two-step framework evaluated on the Genomes to Fields (G2F) 2022 dataset. In Step 1, ML models are employed to fit environmental main effects; in Step 2, genomic residuals are modeled via additive-dominance relationship matrices, augmented by an explicit low-rank G×E matrix. Candidate interaction markers were screened through plasticity-based genome-wide association studies (GWAS) across six phenotypic stability metrics and used to construct a low-rank candidate G×E representation, with a cross-validation-selected scaling parameter applied to control the contribution of the predicted G×E component. TwoStep_G×E_alpha0.33, achieved a within–environment Pearson correlation coefficient (PCC) of 0.376, outperformed both GBLUP and the competition-winning model (PCC = 0.357) in within-environment ranking. Furthermore, environment-adaptive selection yielded a genetic gain of 0.454 Mg ha⁻¹, representing a 34.7% improvement over GBLUP. Overall, the proposed framework provides a practical approach for environment-specific yield prediction and adaptive selection in maize breeding.

Accurate patent valuation remains a persistent challenge in intellectual property management, particularly in the food industry, where technological homogeneity and rapid innovation cycles introduce substantial noise into observable performance indicators. Traditional valuation approaches, whether based on subjective expert judgment or citation-based metrics, often struggle to effectively reduce information uncertainty in this context. To address this limitation, this study proposes an objective, data-driven patent valuation framework grounded in information theory. We construct a multidimensional evaluation system comprising nine indicators across technological, legal, and economic dimensions and apply it to a large-scale dataset of 100,648 invention patents. To address the heavy-tailed nature of patent indicators without sacrificing the information contained in high-impact outliers, we introduce a square-root transformation strategy that stabilizes dispersion while preserving ordinal relationships. Indicator weights are determined objectively via Shannon entropy, capturing the relative scarcity and discriminatory information content of each signal, after which comprehensive value scores are derived using the TOPSIS method. Empirical results reveal that the entropy-based model assigns dominant weights to so-called “costly signals”, specifically PCT applications (29.53%) and patent transfers (24.36%). Statistical correlation analysis confirms that these selected indicators are significantly associated with patent value ($p<0.001$), while bootstrapping tests demonstrate the robustness of the resulting weight structure. The model’s validity is further evaluated using an external benchmark (“ground truth”) dataset comprising 55 patents recognized by the China Patent Award. The proposed framework demonstrates substantially stronger discriminatory capability than baseline methods, awarded patents achieve an average score 2.64 times higher than that of ordinary patents, and the enrichment factor for award-winning patents within the Top-100 ranking reaches 91.5. Additional robustness analyses, including benchmarking against the Weighted Sum Model (WSM), further confirm the methodological stability of the framework, with sensitivity analysis revealing an exceptional enrichment factor of 183.1 for the Top-50 patents. These findings confirm that the Entropy–TOPSIS framework functions as an effective information-filtering mechanism, amplifying high-value patent signals in noise-intensive environments. Consequently, the proposed model serves as a generalizable and theoretically grounded tool for objective patent valuation, with particular relevance to industries characterized by heavy-tailed data and high information uncertainty.