[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-3460":3,"related-3460":63},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":6,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":24,"tags":26,"search_phrases":33,"slug":36,"view_count":37,"doi":38,"paper":39,"created_at":62},3460,"Knowledge graph-based expert system for tea pesticide residue compliance verification and detection method selection optimization","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffsufs.2026.1865601","Introduction Strict Maximum Residue Limits (MRLs) and complex residue definitions in pesticide standards (e.g., GB 2763) pose significant challenges for compliance verification and detection method selection in tea safety management. Manual inspection is prone to systematic errors, particularly false negatives caused by neglecting metabolites or isomers. Methods This study proposes a Knowledge Graph-based Expert System to address these issues. We constructed a domain-specific knowledge graph integrating 110 tea-registered pesticide active substances (comprising 847 entities and 2,156 relationships) and 37 detection standards using a human-in-the-loop verification process. A semantic classification model categorizing residue definitions into six patterns (Type A–F) was developed to enable automated, logic-driven compliance verification. A weighted set-cover algorithm was implemented to optimize the selection of detection methods among existing standardized methods. Validation included a small exploratory pilot user study (n = 5) and a retrospective analysis of 500 historical laboratory samples. Results In the retrospective analysis, the system’s automated verdicts agreed with the laboratory’s human Final Verdict in all 500 cases (95% CI: 99.3%–100.0%), demonstrating faithful reproduction of institutional interpretation of GB 2763. Notably, the system independently flagged three metabolite-summation (Type B) cases that first-level manual review had passed, illustrating its value as a redundant safeguard for multi-component residue definitions. The pilot user study illustrated qualitative trends consistent with cognitive load theory, whereas the system produced deterministic verdicts at a mean of 0.63 s per sample. Discussion This system provides an effective decision-support tool for laboratory quality control and regulatory compliance.","引言 农药标准（如GB 2763）中严格的最高残留限量（MRLs）和复杂的残留物定义，给茶叶安全管理中的合规性验证和检测方法选择带来了重大挑战。人工审查容易产生系统性错误，尤其是因忽视代谢物或异构体而导致的假阴性。方法 本研究提出了一种基于知识图谱的专家系统来解决这些问题。我们采用人在回路验证流程，构建了整合110种茶叶登记农药活性物质（包含847个实体和2，156个关系）及37项检测标准的领域专用知识图谱。开发了将残留物定义归类为六种模式（A型–F型）的语义分类模型，以实现自动化、逻辑驱动的合规性验证。实现了加权集合覆盖算法，以在现有标准化方法中优化检测方法的选择。验证包括一项小型探索性试点用户研究（n = 5）和对500份历史实验室样本的回顾性分析。结果 在回顾性分析中，系统的自动判定在所有500个案例中均与实验室人工最终判定一致（95% CI：99.3%–100.0%），表明系统忠实再现了GB 2763的机构解释。值得注意的是，系统独立标记出3例一级人工审查已通过的代谢物加和（B型）案例，说明其作为多组分残留物定义冗余保障的价值。试点用户研究显示了与认知负荷理论一致的定性趋势，而系统以平均每样本0.63秒的速度产生确定性判定。讨论 该系统为实验室质量控制和法规合规提供了有效的决策支持工具。",null,"Frontiers in Sustainable Food Systems","2026-09-24T00:00:00Z","论文",10,false,78,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},16,22,18,13,9,1,"知识图谱专家系统用于茶叶农残合规判定与检测方法优选，500例回溯验证一致率100%，方法新颖、数据扎实，对农产品质量安全监管有实用价值。",[25],{"name":10,"url":6},[27,28,29,30,31,32],"智慧农业","农业人工智能","知识图谱","农药残留","茶叶安全","检测标准",[34,35],"农药残留 知识图谱","GB 2763 茶叶 检测方法","农药残留知识图谱-3460",0,"10.3389\u002Ffsufs.2026.1865601",{"doi":38,"openalex_id":40,"authors":41,"venue":10,"cited_by_count":37,"oa_url":6,"card":55,"direction":59,"ingested_from":61},"W7214240413",[42,44,46,48,50,52],{"name":43,"orcid":9},"Yanyan Zhang",{"name":45,"orcid":9},"Jianrong Wen",{"name":47,"orcid":9},"Xuexiang Su",{"name":49,"orcid":9},"Tingxin Chen",{"name":51,"orcid":9},"Min Xie",{"name":53,"orcid":54},"Jindong Chen","https:\u002F\u002Forcid.org\u002F0000-0002-5560-1431",{"tldr":56,"method":57,"finding":58,"direction":59,"opportunity":60},"构建知识图谱专家系统，自动核验茶叶农药残留合规并优化检测方法选择。","整合110种农药与37项标准构建知识图谱，用语义分类和加权集合覆盖算法。","500例回溯判定与人工一致，并额外发现3例被人工漏检的代谢物加和案例。","农业人工智能与决策模型","可扩展至其他作物与多残留定义，并探索知识图谱与实验室LIMS实时联动。","openalex","2026-09-25T23:30:06.985753Z",{"total":64,"page":22,"page_size":64,"items":65},6,[66,113,147,188,228,278],{"id":67,"title":68,"url":69,"summary":70,"summary_zh":71,"content":9,"source_name":72,"source_url":69,"published_at":73,"category":12,"cover_url":9,"hotness":13,"is_selected":74,"score":75,"score_detail":76,"sources":81,"tags":83,"search_phrases":86,"slug":89,"view_count":90,"doi":91,"paper":92,"created_at":112},1764,"Human-Centric, AI-Enabled Concurrent Engineering: Towards a cross-domain framework","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.jii.2026.101192","The convergence of Artificial Intelligence (AI) and Concurrent Engineering (CE) has begun to reshape how industrial systems are designed, validated, and operated. We argue that the next, and arguably more consequential, step is to embed this convergence inside a coherent Industrial Information Integration (III) substrate that is explicitly human-centric and that generalises beyond the classical manufacturing core. This paper develops such a framework. We propose the H-AICE architecture (Human-centric AI-enabled Concurrent Engineering), which couples (i) a five-dimensional taxonomy of AI interventions in CE, (ii) a time-dependent overall-risk model Roverall(t) that integrates internal and external risk along the design lifecycle, and (iii) an integration layer based on shared ontologies, knowledge graphs, and digital twins. The framework is then instantiated and critically discussed in six application scenarios that exhibit very different data regimes, regulatory contexts, and human stakes: precision agriculture (Agriculture 5.0), biotechnology and life sciences, climate-aware industrial sustainability, sustainable forestry (Forestry 5.0), medical device development, and livestock breeding. We show that the same III primitives—data quality, semantic interoperability, explainable models, and human-in-the-loop oversight—recur across all six scenarios, and that the principal failure modes are organisational rather than algorithmic. We conclude with a research agenda for human-centric, AI-enabled CE that emphasises explainability, governance, and sector-specific certification, and that positions III as a load-bearing rather than a cosmetic discipline.","人工智能（AI）与并行工程（CE）的融合已开始重塑工业系统的设计、验证与运行方式。我们认为，下一步——或许也是更具深远意义的一步——是将这种融合嵌入一个连贯的工业信息集成（III）基底之中，该基底明确以人为中心，并超越传统制造业核心进行泛化。本文构建了这样一个框架。我们提出H-AICE架构（以人为中心的AI赋能并行工程），该架构结合了（i）AI在CE中干预措施的五维分类法，（ii）一个随时间变化的总体风险模型Roverall(t)，该模型沿设计生命周期整合内部与外部风险，以及（iii）基于共享本体、知识图谱和数字孪生的集成层。随后，该框架在六个应用场景中得到实例化并加以批判性讨论，这些场景展现出截然不同的数据体制、监管环境及人类利害关系：精准农业（农业5.0）、生物技术与生命科学、气候感知型工业可持续性、可持续林业（林业5.0）、医疗器械开发以及畜牧育种。我们表明，相同的III原语——数据质量、语义互操作性、可解释模型及人在回路监督——在全部六个场景中反复出现，且主要失效模式属于组织层面而非算法层面。最后，我们提出一个以人为中心的AI赋能CE研究议程，该议程强调可解释性、治理机制及行业特定认证，并将III定位为一门承重学科而非装饰性学科。","Journal of Industrial Information Integration","2026-09-03T00:00:00Z",true,79,{"impact":77,"substance":18,"depth":19,"authority":78,"freshness":79,"relevant":22,"comment":80},20,14,5,"提出跨领域人本AI并发工程框架，覆盖农业5.0等场景，具前瞻性与方法论价值。",[82],{"name":72,"url":69},[84,27,28,29,85],"数字乡村","农业5.0",[87,88],"农业人工智能 数字乡村 智慧农业 知识图谱","农业人工智能 数字乡村","农业人工智能数字乡村智慧农业知识图谱-1764",2,"10.1016\u002Fj.jii.2026.101192",{"doi":91,"openalex_id":93,"authors":94,"venue":72,"cited_by_count":37,"oa_url":104,"card":105,"direction":111,"ingested_from":61},"W7167326837",[95,98,101],{"name":96,"orcid":97},"Andreas Holzinger","https:\u002F\u002Forcid.org\u002F0000-0002-6786-5194",{"name":99,"orcid":100},"Heimo Müller","https:\u002F\u002Forcid.org\u002F0000-0002-9691-4872",{"name":102,"orcid":103},"Josip Stjepandić","https:\u002F\u002Forcid.org\u002F0000-0003-0877-3799","https:\u002F\u002Fwww.sciencedirect.com\u002Fscience\u002Farticle\u002Fpii\u002FS2452414X26001342\u002Fpdf",{"tldr":106,"method":107,"finding":108,"direction":109,"opportunity":110},"提出人本AI并发工程框架，跨领域应用于农业等六场景。","提出H-AICE架构，结合AI干预分类、风险模型、本体与数字孪生。","跨场景共性为数据质量、语义互操作、可解释模型和人机协同，主要失败模式是组织性的。","其他","农业5.0场景中，可研究人机协同决策与组织治理对AI采纳的影响，以及可解释模型在精准农业中的实际应用。","智慧农业 \u002F 农业物联网","2026-09-06T23:30:12.843367Z",{"id":114,"title":115,"url":116,"summary":117,"summary_zh":118,"content":9,"source_name":119,"source_url":116,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":120,"sources":122,"tags":124,"search_phrases":128,"slug":131,"view_count":37,"doi":132,"paper":133,"created_at":146},3517,"A Resource-Efficient Hybrid CNN-LSTM Network for Image-Based Bean Leaf Disease Classification","https:\u002F\u002Fdoi.org\u002F10.3390\u002Fjimaging12100468","Accurate and resource-efficient automated diagnosis is a cornerstone of modern agricultural expert systems. While Convolutional Neural Networks (CNNs) have established benchmarks in plant pathology, their ability to capture long-range spatial dependencies is often limited by standard pooling layers, and their high memory footprint hinders deployment on portable devices. This paper proposes a lightweight hybrid CNN-LSTM system for bean leaf disease classification. By integrating an LSTM layer to model the spatial–sequential relationships within feature maps, our hybrid architecture achieves a 94.36% accuracy and 94.38% F1 score while maintaining an exceptionally small footprint of 1.86 MB, a 70% reduction in size compared to traditional CNN-based systems. Furthermore, we provide a systematic evaluation of image augmentation strategies, demonstrating that tailored transformations are superior to generic combinations for maintaining the integrity of diagnostic patterns. Results on the ibean dataset confirm that the proposed system achieves state-of-the-art F1 scores of 99.22% with EfficientNet-B7+LSTM, providing a potentially robust and scalable framework for real-time agricultural decision support in resource-constrained environments. The code and augmented datasets used in this study are publicly available on this GitHub repo.","准确且资源高效的自动化诊断是现代农业专家系统的基石。尽管卷积神经网络（CNN）在植物病理学领域已确立了基准，但其捕捉长程空间依赖关系的能力常受限于标准池化层，且高内存占用阻碍了其在便携设备上的部署。本文提出了一种用于豆叶病害分类的轻量级混合CNN-LSTM系统。通过集成LSTM层来建模特征图内的空间-序列关系，我们的混合架构达到了94.36%的准确率和94.38%的F1分数，同时保持了仅1.86 MB的极小占用，相较于传统基于CNN的系统体积减少了70%。此外，我们系统评估了图像增强策略，表明定制化变换在保持诊断模式完整性方面优于通用组合。在ibean数据集上的结果证实，所提出的系统结合EfficientNet-B7+LSTM达到了99.22%的最先进F1分数，为资源受限环境中的实时农业决策支持提供了一个潜在稳健且可扩展的框架。本研究使用的代码和增强数据集已在此GitHub仓库公开。","Journal of Imaging",{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":121},"轻量级CNN-LSTM豆叶病害识别研究，方法新颖、数据可靠且代码开源，对资源受限场景下的农业智能诊断有实用价值。",[123],{"name":119,"url":116},[27,28,125,126,127],"病害识别","轻量化模型","豆类作物",[129,130],"ibean dataset 豆叶病害","农业人工智能 轻量化模型 智慧农业 病害识别","ibeandataset豆叶病害-3517","10.3390\u002Fjimaging12100468",{"doi":132,"openalex_id":134,"authors":135,"venue":119,"cited_by_count":37,"oa_url":116,"card":141,"direction":59,"ingested_from":61},"W7154572440",[136,138],{"name":137,"orcid":9},"Hye Jin Rhee",{"name":139,"orcid":140},"Joseph Damilola Akinyemi","https:\u002F\u002Forcid.org\u002F0000-0003-3121-4231",{"tldr":142,"method":143,"finding":144,"direction":59,"opportunity":145},"提出轻量级CNN-LSTM混合网络，用于豆叶病害分类，兼顾高精度与低资源占用。","CNN提取特征后接LSTM建模空间序列关系，在ibean数据集上评估并系统比较图","模型准确率94.36%、F1 94.38%，仅1.86MB，比传统CNN缩小70%，Efficien","可探索面向移动端\u002F边缘设备的超轻量病害诊断模型，并研究增强策略与模型结构的自适应协同优化。","2026-09-25T23:30:59.276381Z",{"id":148,"title":149,"url":150,"summary":151,"summary_zh":152,"content":9,"source_name":153,"source_url":150,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":154,"score_detail":155,"sources":160,"tags":162,"search_phrases":166,"slug":169,"view_count":37,"doi":170,"paper":171,"created_at":187},3515,"Soil mapping and fertilizer optimization for precision agriculture using artificial intelligence","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs13198-026-03435-1","Soil mapping and fertilizer optimization for precision agriculture using artificial intelligence。International Journal of Systems Assurance Engineering and Management","基于人工智能的精准农业土壤制图与肥料优化。《国际系统保障工程与管理杂志》","International Journal of Systems Assurance Engineering and Management",62,{"impact":156,"substance":78,"depth":157,"authority":20,"freshness":158,"relevant":22,"comment":159},12,15,8,"论文探讨AI用于土壤制图与施肥优化，属智慧农业细分方向，但摘要信息有限、影响面偏窄，暂不建议进入每日精选。",[161],{"name":153,"url":150},[27,28,163,164,165],"精准施肥","遥感","土壤制图",[167,168],"土壤制图 人工智能 精准施肥","精准农业 肥料优化 AI","土壤制图人工智能精准施肥-3515","10.1007\u002Fs13198-026-03435-1",{"doi":170,"openalex_id":172,"authors":173,"venue":153,"cited_by_count":37,"oa_url":9,"card":182,"direction":59,"ingested_from":61},"W7214144821",[174,177,179],{"name":175,"orcid":176},"Neetu Mittal","https:\u002F\u002Forcid.org\u002F0000-0002-2012-0523",{"name":175,"orcid":178},"https:\u002F\u002Forcid.org\u002F0000-0001-6923-0013",{"name":180,"orcid":181},"Pradeepta Kumar Sarangi","https:\u002F\u002Forcid.org\u002F0000-0003-3827-6208",{"tldr":183,"method":184,"finding":185,"direction":59,"opportunity":186},"利用人工智能进行土壤制图和肥料优化，以支持精准农业。","人工智能方法，用于土壤制图与肥料优化。","AI可提升土壤制图与肥料优化的精准性，促进精准农业。","可探索多源数据融合与实时决策模型，提升肥料推荐的自适应性和可解释性。","2026-09-25T23:30:49.869002Z",{"id":189,"title":190,"url":191,"summary":192,"summary_zh":193,"content":9,"source_name":194,"source_url":191,"published_at":195,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":196,"score_detail":197,"sources":201,"tags":203,"search_phrases":207,"slug":210,"view_count":37,"doi":211,"paper":212,"created_at":227},3514,"Artificial Intelligence for Climate Adaptation Decision Support in Data-Poor Developing Regions","https:\u002F\u002Fdoi.org\u002F10.22541\u002Fessoar.15009304\u002Fv1","Climate adaptation is a sequence of decisions taken under uncertainty, and the regions where climate risk is rising fastest are those with the least information to guide them. Only about 10 per cent of deaths are registered in the WHO African Region; nearly 90 per cent of required surface weather observations are missing across least developed countries and small island states; and only 40 per cent of African countries have multi-hazard early warning systems. This report examines whether artificial intelligence — machine learning, remote sensing and predictive analytics — can close these information gaps and improve adaptation decisions in data-poor developing regions. The report organises the problem as a decision chain with three information gaps — observation, prediction and decision — followed by an action gap that AI cannot close. It finds that AI has advanced fastest on prediction: AI weather models became operational at ECMWF in 2025, AI flood forecasts now cover 100 countries and about 700 million people, satellite nowcasts reach a continent with little radar, and AI monsoon-onset forecasts reached 38 million Indian farmers in 2025. On observation, satellite machine learning explains around 70 per cent of the variation in village wealth but only up to about half of the variation in changes over time. On decision, evidence from Togo, Bangladesh and Kenya shows that AI-assisted targeting, forecast-based triggers and satellite index insurance can deliver assistance faster and better, within clear limits. The report's central argument is the ground-truth paradox: AI stretches scarce observations further, but every AI product must be trained and verified against ground truth, so reliance on AI raises the value of each remaining station, survey and label. The 2025 interruption of FEWS NET and termination of the DHS Program show how fragile that foundation is. Because the value of information is the product of skill, lead time, reach, trust and the means to act, the highest returns usually lie not in more skilful models but in dissemination, institutions and prearranged finance. The report sets out a risk register, a six-principle policy framework, actions by actor and a roadmap to 2030.","气候适应是在不确定性下做出的一系列决策，而气候风险上升最快的地区恰恰是指导信息最匮乏的地区。世卫组织非洲区域仅登记了约10%的死亡病例；最不发达国家和小岛屿国家缺失了近90%所需的地面天气观测数据；仅有40%的非洲国家拥有多灾种早期预警系统。本报告考察人工智能——机器学习、遥感和预测分析——能否弥合这些信息缺口，改善数据匮乏的发展中地区的适应决策。报告将这一问题组织为一条决策链，包含三个信息缺口——观测、预测和决策——以及一个人工智能无法弥合的行动缺口。报告发现，人工智能在预测方面进展最快：人工智能天气模型于2025年在欧洲中期天气预报中心（ECMWF）投入业务运行，人工智能洪水预报现已覆盖100个国家和约7亿人口，卫星临近预报覆盖了一个几乎没有雷达的大陆，人工智能季风爆发预报于2025年惠及3800万印度农民。在观测方面，卫星机器学习可解释村庄财富约70%的变异，但对时间变化的解释力仅约一半。在决策方面，来自多哥、孟加拉国和肯尼亚的证据表明，人工智能辅助的目标定位、基于预报的触发机制和卫星指数保险能够在明确限度内更快、更好地提供援助。报告的核心论点是地面真值悖论：人工智能能够将稀缺的观测数据发挥更大效用，但每个人工智能产品都必须依据地面真值进行训练和验证，因此对人工智能的依赖提升了每一个剩余站点、调查和标注数据的价值。2025年FEWS NET的中断和DHS项目的终止表明这一基础何等脆弱。由于信息的价值是技能、提前期、覆盖面、信任和行动手段的乘积，最高回报通常不在于更精密的模型，而在于传播、制度和预先安排的融资。报告提出了风险登记册、六项原则的政策框架、各行为主体的行动以及到2030年的路线图。","OpenAlex","2026-09-22T00:00:00Z",86,{"impact":18,"substance":198,"depth":199,"authority":20,"freshness":158,"relevant":22,"comment":200},24,19,"系统梳理AI在数据匮乏地区气候适应决策中的观测、预测与决策三类信息缺口，提出“地面真值悖论”，数据与结论扎实，对农业信息化与智慧农业有较强参考价值。",[202],{"name":194,"url":191},[27,28,204,205,206],"气候适应","遥感监测","早期预警",[208,209],"AI 气候适应 决策支持","数据匮乏地区 农业预警","AI气候适应决策支持-3514","10.22541\u002Fessoar.15009304\u002Fv1",{"doi":211,"openalex_id":213,"authors":214,"venue":9,"cited_by_count":37,"oa_url":220,"card":221,"direction":226,"ingested_from":61},"W7214097088",[215,217],{"name":216,"orcid":9},"H Heuristics",{"name":218,"orcid":219},"Hunter Hughes","https:\u002F\u002Forcid.org\u002F0009-0002-6161-9387","https:\u002F\u002Fessopenarchive.org\u002Fdoi\u002Fpdf\u002F10.22541\u002Fessoar.15009304\u002Fv1",{"tldr":222,"method":223,"finding":224,"direction":59,"opportunity":225},"评估AI能否弥补数据匮乏地区气候适应决策的信息缺口，并提出地面真值悖论。","梳理观测、预测、决策三环节，结合AI天气模型、卫星ML与多国案例证据。","AI预测进展最快，但依赖地面真值；最高回报常在传播、制度与预置资金而非模型。","可研究AI辅助农业气候适应中地面真值稀缺下的验证与信任机制，及预置资金触发设计。","数字乡村与农业信息化","2026-09-25T23:30:46.008325Z",{"id":229,"title":230,"url":231,"summary":232,"summary_zh":233,"content":9,"source_name":234,"source_url":231,"published_at":235,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":236,"score_detail":237,"sources":239,"tags":241,"search_phrases":245,"slug":248,"view_count":37,"doi":249,"paper":250,"created_at":277},3513,"Accelerated development and deployment of computer vision models for invasive aquatic pests","https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs41598-026-72212-8","Abstract Aquatic non-indigenous species (NIS) incur significant cultural, environmental and economic costs worldwide, and human surveillance is expensive and impractical at large scales. Computer vision models (CVMs) deployed on remote and autonomous vehicles can reduce the burden on trained human observers, but aquatic environments present unique challenges and lack frameworks for biosecurity-focused development and deployment. We present a practical framework for developing and deploying new species-specific CVMs for real-time use across surface vessel and remote vehicle platforms. We demonstrate the utility of our framework through its application to several taxonomically and ecologically diverse NIS in New Zealand: Mediterranean fanworm Sabella spallanzanii , South African oxygen weed Lagarosiphon major, and exotic Caulerpa ( Caulerpa brachypus and C. parvifolia ). For S. spallanzanii , we demonstrate efficient training from small datasets and deployment for real-time detection. Using exotic Caulerpa , we show how CVMs can be rapidly improved during an early incursion response. For L. major , standardised field validation methods enable the comparison of CVM and human detection rates across diverse locations and operating conditions. Collectively, these case studies demonstrate that our framework enables accurate detection and the robust assessment of model effectiveness under realistic field conditions and can be effectively applied to imagery from multiple platforms.","摘要 水生外来非本土物种（NIS）在全球范围内造成显著的文化、环境和经济损失，而人工监测成本高昂且难以大规模实施。部署在远程和自主载具上的计算机视觉模型（CVM）可减轻对训练有素的人类观察者的依赖，但水生环境面临独特挑战，且缺乏以生物安全为重点的开发和部署框架。我们提出了一个实用框架，用于开发和部署新的物种特异性CVM，以在水面船只和远程载具平台上实时使用。我们通过将该框架应用于新西兰多个在分类学和生态学上具有多样性的NIS来展示其效用：地中海缨鳃虫 Sabella spallanzanii、南非氧草 Lagarosiphon major 以及外来Caulerpa（Caulerpa brachypus 和 C. parvifolia）。对于 S. spallanzanii，我们展示了从小型数据集进行高效训练并部署用于实时检测。利用外来Caulerpa，我们展示了CVM如何在入侵早期响应期间快速改进。对于 L. major，标准化野外验证方法使得能够在不同地点和操作条件下比较CVM与人类检测率。总体而言，这些案例研究表明，我们的框架能够在现实野外条件下实现准确检测和对模型有效性的稳健评估，并可有效应用于来自多个平台的图像。","Scientific Reports","2026-09-23T00:00:00Z",80,{"impact":19,"substance":18,"depth":19,"authority":78,"freshness":158,"relevant":22,"comment":238},"提出面向水生入侵物种的计算机视觉模型开发与部署框架，含多物种实地验证，方法新颖且具生物安全应用价值，但属新西兰案例，对国内农业信息化影响有限。",[240],{"name":234,"url":231},[27,28,242,243,244],"计算机视觉","入侵物种监测","水生生物安全",[246,247],"新西兰 入侵水生物种 计算机视觉","Caulerpa 实时检测 模型","新西兰入侵水生物种计算机视觉-3513","10.1038\u002Fs41598-026-72212-8",{"doi":249,"openalex_id":251,"authors":252,"venue":234,"cited_by_count":37,"oa_url":271,"card":272,"direction":226,"ingested_from":61},"W7214089404",[253,256,258,260,262,265,268],{"name":254,"orcid":255},"Rose A. Pearson","https:\u002F\u002Forcid.org\u002F0000-0002-4700-2113",{"name":257,"orcid":9},"Gareth Preston",{"name":259,"orcid":9},"Jeremy Bulleid",{"name":261,"orcid":9},"Svenja David",{"name":263,"orcid":264},"Felix Vaux","https:\u002F\u002Forcid.org\u002F0000-0002-2882-7996",{"name":266,"orcid":267},"Daniel Clements","https:\u002F\u002Forcid.org\u002F0000-0001-9319-5588",{"name":269,"orcid":270},"Leigh W. Tait","https:\u002F\u002Forcid.org\u002F0000-0001-9153-139X","https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41598-026-72212-8_reference.pdf",{"tldr":273,"method":274,"finding":275,"direction":59,"opportunity":276},"提出一套开发与部署水生入侵物种计算机视觉模型的实用框架，并在新西兰三种入侵物种上验证。","基于水面船与遥控平台图像，小数据集训练、实时检测与标准化野外验证。","框架能实现实时准确检测，并在真实野外条件下稳健评估模型效果。","可迁移至农业入侵生物监测，探索小样本跨平台模型与野外标准化评估体系。","2026-09-25T23:30:42.003495Z",{"id":279,"title":280,"url":281,"summary":282,"summary_zh":283,"content":9,"source_name":284,"source_url":281,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":236,"score_detail":285,"sources":287,"tags":289,"search_phrases":292,"slug":295,"view_count":37,"doi":296,"paper":297,"created_at":310},3512,"AI-Driven Agricultural Advisory and Diagnostic Systems for Smallholder Farming: Technical Architectures, Evidence and Deployment Priorities for North-East India","https:\u002F\u002Fdoi.org\u002F10.9734\u002Farja\u002F2026\u002Fv19i4919","Artificial intelligence (AI) is being introduced into agricultural advisory services through machine learning, computer vision, conversational large language models, retrieval-augmented generation and multimodal interfaces. For smallholder farming, the central question is not whether these technologies can produce technically plausible outputs, but whether they can provide locally correct, actionable and safe recommendations under heterogeneous agronomic, linguistic and connectivity conditions. This critical narrative review integrates evidence on digital extension, AI-enabled agricultural advice, image-based diagnosis and responsible digital agriculture, with particular reference to North-East India. Literature published from 1 January 2010 to 17 July 2026 was considered, with emphasis on peer-reviewed field evaluations, technical validation studies, reviews and regionally relevant research. Evidence from digital extension provides the strongest causal baseline: mobile and personalised advisory services frequently improve information recall, agronomic knowledge and adoption of recommended practices, yet effects on yield, profit and welfare are inconsistent. Recent generative-AI studies show that large language models can produce useful agricultural responses, but site-specific rates, timing and local practice remain recurrent failure points. Retrieval grounding and expert feedback improve local relevance, although multi-season farm-level effectiveness evidence remains scarce. Image-based plant-disease systems achieve high accuracy in curated datasets, but performance can deteriorate sharply under field domain shift, class novelty and variable image quality. North-East Indian studies of mobile advisory systems in Meghalaya, Nagaland and Tripura demonstrate a valuable institutional foundation based on interactive voice response, local expert networks and user-centred service design; they do not, however, establish the effectiveness of autonomous AI. The most defensible deployment model is therefore an offline-tolerant, multilingual, multimodal and human-supervised architecture that grounds recommendations in curated regional knowledge, represents uncertainty, preserves provenance and escalates high-risk or out-of-distribution cases. Future research should prioritise prospective district- and season-spanning evaluations that connect model quality to farmer decisions, agronomic outcomes, equity, safety and cost-effectiveness.","人工智能（AI）正通过机器学习、计算机视觉、对话式大语言模型、检索增强生成和多模态界面被引入农业咨询服务。对于小农户而言，核心问题不在于这些技术能否产生技术上看似合理的输出，而在于它们能否在异质的农艺、语言和网络连接条件下提供本地正确、可操作且安全的建议。本批判性叙事综述整合了数字推广、AI赋能的农业建议、基于图像的诊断和负责任数字农业方面的证据，并特别关注印度东北部。本文考察了2010年1月1日至2026年7月17日期间发表的文献，重点关注同行评议的田间评估、技术验证研究、综述及区域相关研究。来自数字推广的证据提供了最强的因果基线：移动化和个性化咨询服务经常改善信息记忆、农艺知识和对推荐措施的采纳，但对产量、利润和福利的影响并不一致。近期生成式AI研究表明，大语言模型能够产生有用的农业回答，但针对具体地点的用量、时机和本地实践仍是反复出现的失败点。检索 grounding 和专家反馈可提高本地相关性，但多季农场层面的有效性证据仍然稀缺。基于图像的植物病害系统在精选数据集上达到高准确率，但在田间域偏移、类别新颖性和图像质量多变的情况下，性能可能急剧下降。印度东北部在梅加拉亚邦、那加兰邦和特里普拉邦开展的移动咨询系统研究展示了基于交互式语音应答、本地专家网络和以用户为中心的服务设计的宝贵制度基础；然而，这些研究并未确立自主AI的有效性。因此，最可辩护的部署模式是一种容忍离线、多语言、多模态且有人工监督的架构，该架构将建议建立在精选的区域知识之上，表征不确定性，保留来源信息，并对高风险或分布外案例进行升级处理。未来研究应优先开展前瞻性的跨区县和跨季节评估，将模型质量与农户决策、农艺结果、公平性、安全性和成本效益联系起来。","Asian Research Journal of Agriculture",{"impact":19,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":286},"系统综述AI农业咨询与诊断系统在印度东北小农场景的技术架构与落地证据，指出人机协同、离线多语言与检索增强是可行路径，对智慧农业落地有参考价值。",[288],{"name":284,"url":281},[84,27,28,290,291],"农业技术推广","小农户",[293,294],"印度东北部 农业AI 小农户","农业智能诊断 多语言 离线","印度东北部农业AI小农户-3512","10.9734\u002Farja\u002F2026\u002Fv19i4919",{"doi":296,"openalex_id":298,"authors":299,"venue":284,"cited_by_count":37,"oa_url":281,"card":305,"direction":226,"ingested_from":61},"W7214205238",[300,302],{"name":301,"orcid":9},"Pravangkar Boruah",{"name":303,"orcid":304},"Rubul Kumar Bania","https:\u002F\u002Forcid.org\u002F0000-0001-6294-0231",{"tldr":306,"method":307,"finding":308,"direction":59,"opportunity":309},"综述AI农业咨询与诊断系统，聚焦印度东北小农，提出人监督多模态部署架构。","批判性叙述综述，整合2010-2026年数字推广、生成式AI与图像诊断证据。","AI输出技术可行但本地化、安全与田间效果证据不足，需人监督与检索增强。","可开展跨区跨季前瞻评估，连接模型质量与农户决策、产量、公平及成本效益。","2026-09-25T23:30:39.745514Z"]