Artificial Intelligence in Agriculture最新文献

筛选
英文 中文
From on-device vision to assisted harvesting: A lightweight real-time ripeness evaluation system for truss-harvested cherry tomatoes on augmented reality glasses 从设备上的视觉到辅助收获:一个轻量级的实时成熟度评估系统,用于增强现实眼镜上的桁架收获的圣女果
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-08-10 DOI: 10.1016/j.aiia.2026.08.003
Yang Shao, Jiale Cui, Chen Tang, Zhenqing Liu, Fengjiao Xu, Ruifang Zhai, Wanneng Yang, Peng Song
{"title":"From on-device vision to assisted harvesting: A lightweight real-time ripeness evaluation system for truss-harvested cherry tomatoes on augmented reality glasses","authors":"Yang Shao,&nbsp;Jiale Cui,&nbsp;Chen Tang,&nbsp;Zhenqing Liu,&nbsp;Fengjiao Xu,&nbsp;Ruifang Zhai,&nbsp;Wanneng Yang,&nbsp;Peng Song","doi":"10.1016/j.aiia.2026.08.003","DOIUrl":"10.1016/j.aiia.2026.08.003","url":null,"abstract":"<div><div>The transition from experience-driven manual harvesting to data-driven intelligent management represents a pivotal shift in modern facility agriculture. However, the widespread adoption of fully autonomous systems remains constrained by the environmental complexity of greenhouses and the high cost of agricultural robotics. Focusing on the practical challenges of truss-harvested tomato production, this study proposes an augmented reality (AR)-assisted real-time maturity assessment and in-situ decision-support system that bridges manual labor and intelligent automation through human–machine collaboration. At the core of the system is DGTruss-YOLO, an improved lightweight YOLOv11 model tailored for resource-constrained wearable devices. The architecture integrates a Spatial–Channel Reconstruction Convolution (ScConv) module to suppress spatial and channel-wise feature redundancy, and a Multi-scale Feature Modulation-based Auxiliary Feature Pyramid Network (MFM-MAFPN) to jointly capture fine-grained individual fruit textures and the global structural characteristics of tomato trusses. To ensure seamless edge deployment, a simplified standard convolution-based Efficient_Detect head is employed to enhance hardware compatibility and inference efficiency on AR smart glasses. Furthermore, a practical weighted grading strategy is introduced to map fruit-level recognition results to a five-grade (A-E) truss maturity index, aligning visual perception outcomes with practical harvesting requirements. Experimental results demonstrate that the optimized model reduces parameter count and computational complexity by 11.07% and 12.70%, respectively, compared to the baseline YOLOv11, while maintaining a stable real-time inference speed of 15–20 FPS on AR smart glasses. The proposed system achieves a truss-level maturity assessment accuracy of 95.26% with 100% adjacent-tolerance accuracy, providing a consistent model-based truss-level maturity assessment to support harvesting decisions in complex greenhouse environments. By providing intuitive, hands-free, and in-situ visual guidance, this work establishes a practical technical pipeline for human–machine collaborative harvesting and offers a practical reference for deploying edge-side deep learning solutions in sustainable and intelligent agricultural production.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 1984-1999"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149347683","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Toward autonomous agriculture: Integrating computer vision and large language models for intelligent weed management 迈向自主农业:集成计算机视觉和大型语言模型的智能杂草管理
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-05-30 DOI: 10.1016/j.aiia.2026.05.002
Yongda Lin, Ruibo Zhang, Renzhong Fu, Yunxiao Wu, Xiaofei Song, Honggen Chang, Shiyu Xia, Yajia Liu
{"title":"Toward autonomous agriculture: Integrating computer vision and large language models for intelligent weed management","authors":"Yongda Lin,&nbsp;Ruibo Zhang,&nbsp;Renzhong Fu,&nbsp;Yunxiao Wu,&nbsp;Xiaofei Song,&nbsp;Honggen Chang,&nbsp;Shiyu Xia,&nbsp;Yajia Liu","doi":"10.1016/j.aiia.2026.05.002","DOIUrl":"10.1016/j.aiia.2026.05.002","url":null,"abstract":"<div><div>Weed diagnosis and management in agricultural fields still rely heavily on manual observation and expert knowledge, resulting in limited efficiency, accessibility, and decision interactivity. Although computer vision enables automated weed detection and large language models (LLMs) provide new possibilities for intelligent decision support, an integrated framework that connects visual perception with management reasoning is still lacking. To address this challenge, this study proposes an integrated framework for intelligent weed diagnosis and management that combines a weed detection model, an LLM-based decision module, and a user-oriented diagnosis system. A detection model named Mamba-Weed is developed based on the Mamba object detection architecture, incorporating a Lightweight Adaptive Extraction (LAE) module and a Semantics and Detail Infusion (SDI) module to enhance discriminative feature learning and multi-scale feature fusion. An intelligent weed management decision module is then constructed using the DeepSeek LLM, together with Retrieval-Augmented Generation (RAG) and structured prompt engineering to generate domain-specific management recommendations. Experiments on the Weed25 dataset show that Mamba-Weed achieves 92.10% precision, 89.50% recall, and 93.80% mAP@50, demonstrating strong detection performance in complex field environments. The proposed framework establishes a unified pipeline from visual weed identification to intelligent decision-making, providing a practical solution for real-time field weed monitoring and scalable precision weed management.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 1438-1456"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149347944","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Hyperspectral–machine learning framework enables early and non-destructive prediction of plant resistance to pest 高光谱-机器学习框架使植物抗虫害的早期和非破坏性预测成为可能
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-08-08 DOI: 10.1016/j.aiia.2026.07.008
Pin-Jun Wan, Qi-Xin Yang, Yu-Biao Cai, Xin-Feng Wang, Ya-Xuan Wang, Feng-Xiang Lai, Qi Wei, Jia-Chun He, Wei-Xia Wang, Jin-Li Zhang, Qiang Fu
{"title":"Hyperspectral–machine learning framework enables early and non-destructive prediction of plant resistance to pest","authors":"Pin-Jun Wan,&nbsp;Qi-Xin Yang,&nbsp;Yu-Biao Cai,&nbsp;Xin-Feng Wang,&nbsp;Ya-Xuan Wang,&nbsp;Feng-Xiang Lai,&nbsp;Qi Wei,&nbsp;Jia-Chun He,&nbsp;Wei-Xia Wang,&nbsp;Jin-Li Zhang,&nbsp;Qiang Fu","doi":"10.1016/j.aiia.2026.07.008","DOIUrl":"10.1016/j.aiia.2026.07.008","url":null,"abstract":"<div><div>The brown planthopper (<em>Nilaparvata lugens</em>) is one of the most destructive pests of rice and poses a threat to yield stability and food security. Although host-plant resistance is the most sustainable strategy for BPH management, conventional resistance phenotyping remains labor-intensive, destructive, and poorly suited for large-scale breeding. Here, we combined hyperspectral reflectance profiling of 50 rice varieties with an interpretable machine learning framework to enable non-destructive prediction of resistance phenotypes. Using post-infestation spectral profiles, we established classification models that captured resistance states shaped by constitutive traits and inducible defense responses. Among 13 evaluated algorithms, a radial basis function support vector machine achieved the best performance on full-spectrum data within the sampled variety panel, with an average accuracy of 0.939 ± 0.015 and a maximum of 0.972. Predictive wavelengths were concentrated in the green, red-edge, and near-infrared regions, corresponding to variation in pigment regulation, canopy structure, and water status. Spectral and network analyses showed that resistant genotypes exhibited more complex but less stable spectral co-occurrence networks, consistent with physiological trade-offs associated with defense. We also tested whether resistance could be predicted before pest infestation. Pre-infestation spectra retained significant predictive power, with accuracies of 0.572 ± 0.021 for five-class classification and 0.667 ± 0.021 for binary classification, indicating that constitutive defense-associated physiological states are optically detectable before visible damage occurs. Together, our results show that hyperspectral reflectance encodes both inducible responses after infestation and constitutive defense baselines present beforehand. This work establishes a scalable, non-invasive phenotyping strategy for early resistance screening within evaluated germplasm panels, while future validation across independent and variety-level held-out populations will be required before broader deployment.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 1899-1912"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149361778","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
YOLO11-FFTDA: fusing frequency-domain enhancement and deformable attention for robust tomato pedicel segmentation YOLO11-FFTDA:融合频域增强和形变注意的稳健番茄花梗分割
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-08-20 DOI: 10.1016/j.aiia.2026.08.007
Chengjin Qin, Yangtian Lin, Hanchen Zheng, Yizhou Yang, Shuo Zhang, Leye Cheng, Haodi Wang, Liang Gong, Chengliang Liu
{"title":"YOLO11-FFTDA: fusing frequency-domain enhancement and deformable attention for robust tomato pedicel segmentation","authors":"Chengjin Qin,&nbsp;Yangtian Lin,&nbsp;Hanchen Zheng,&nbsp;Yizhou Yang,&nbsp;Shuo Zhang,&nbsp;Leye Cheng,&nbsp;Haodi Wang,&nbsp;Liang Gong,&nbsp;Chengliang Liu","doi":"10.1016/j.aiia.2026.08.007","DOIUrl":"10.1016/j.aiia.2026.08.007","url":null,"abstract":"<div><div>Precise detection and segmentation of tomato pedicels are critical perception tasks for automated harvesting by picking robots. Addressing challenges posed by complex orchard environments and loss of small target features, this paper constructs a tomato pedicel dataset and proposes the YOLO11-FFTDA model. First, the backbone network incorporates a CFP module featuring frequency-domain attention (FFTAT) to explicitly enhance high-frequency feature extraction for slender stems. Second, the integrated C2PSADA module employs deformable attention to adaptively fit stem shapes, effectively suppressing background noise and occlusions. Third, a small object detection branch is added to improve capture capability for minute targets. Experimental results demonstrate that the YOLO11-FFTDA model achieves 88.1% precision, 76.6% recall, 80.0% mAP50, and 32.9% mAP50–95. Compared to the baseline YOLO11 model, its mask quality metric mAP50–95 improves by 3.0 percentage points and compared to the lightweight Yoloact model, mAP50 and mAP50–95 increased significantly by 19.9% and 9.8%, respectively. Five-fold cross-validation revealed extremely low standard deviation, validating the model's high stability and robustness against data variance in complex orchard scenarios. This method enables fast and accurate tomato stem segmentation in computationally constrained complex orchard environments, providing reliable technical support for the visual servo system of harvesting robots.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 1405-1427"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148916155","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Integrating machine learning and large language models to enhance risk assessment and priority management for plant biosecurity 整合机器学习和大型语言模型,加强植物生物安全风险评估和优先级管理
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-07-31 DOI: 10.1016/j.aiia.2026.07.004
Haoxiang Zhao, Shanqing Yi, Andy Sheppard, Rieks van Klinken, Nianwan Yang, Jue Wang, Xiaoqing Xian, Fanghao Wan, Wanxue Liu
{"title":"Integrating machine learning and large language models to enhance risk assessment and priority management for plant biosecurity","authors":"Haoxiang Zhao,&nbsp;Shanqing Yi,&nbsp;Andy Sheppard,&nbsp;Rieks van Klinken,&nbsp;Nianwan Yang,&nbsp;Jue Wang,&nbsp;Xiaoqing Xian,&nbsp;Fanghao Wan,&nbsp;Wanxue Liu","doi":"10.1016/j.aiia.2026.07.004","DOIUrl":"10.1016/j.aiia.2026.07.004","url":null,"abstract":"<div><div>Global food security and biosecurity are continually threatened by the prevalence and spread of plant pests with the ever-accelerating rate of global trade. The continuous advancement of artificial intelligence, such as large language models (LLMs) and machine learning, provides novel insights into plant pest risk management in the food trade, whereas their practical effectiveness remains insufficiently understood. Here, by integrating an LLM with the retrieval-augmented generation method, a machine learning-based network analysis and ecological niche modeling, we constructed a risk management framework and a decision support system to assess the multistage risk of pests in the citrus trade, thereby supporting priority risk management. The introduction risk of citrus pests tended to increase from 2003 to 2018, with 243 introduction risk pathways and 98 risk nodes, mainly including scale insects and fruit flies. Our findings highlight the high risk of co-introduction and co-establishment of multiple citrus pests at important citrus planting areas in southern China, as well as the risk of spread into high-latitude regions under climate change. The decision support platform demonstrated reliable citrus pest risk management performance. Additionally, our study provides novel insights into multistage risk management of plant pests with various trade pathways.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 1852-1865"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149347654","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Integrating SIF-derived parameters with three-band vegetation indices for quantifying maize leaf spot severity: a case study under field conditions 整合sif衍生参数与三波段植被指数量化玉米叶斑病严重程度:田间条件下的案例研究
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-08-17 DOI: 10.1016/j.aiia.2026.08.008
Kehui Ren, Binxiang Qian, Anting Guo, Yi Luo, Jian Zhang, Huiqin Ma, Yi Yang, Shujia Zhuang, Kun Wang, Wenjiang Huang
{"title":"Integrating SIF-derived parameters with three-band vegetation indices for quantifying maize leaf spot severity: a case study under field conditions","authors":"Kehui Ren,&nbsp;Binxiang Qian,&nbsp;Anting Guo,&nbsp;Yi Luo,&nbsp;Jian Zhang,&nbsp;Huiqin Ma,&nbsp;Yi Yang,&nbsp;Shujia Zhuang,&nbsp;Kun Wang,&nbsp;Wenjiang Huang","doi":"10.1016/j.aiia.2026.08.008","DOIUrl":"10.1016/j.aiia.2026.08.008","url":null,"abstract":"<div><div>Leaf spot disease, as a pervasive foliar disease, has become a significant factor limiting stable grain production. Rapid, accurate remote sensing of disease status supports precision prevention and control. Sun-induced chlorophyll fluorescence (SIF) is closely coupled to photosynthesis and highly sensitive to disease stress, but the intertwined physiological and non-physiological influences impede mechanistic attribution and robust quantification of SIF variability, and SIF alone cannot represent the full cascade of stress damage. To overcome these constraints, SIF was decoupled using three reflectance-based approaches (NIRvP, FCVI×PAR, NIRvR) and radiative transfer model simulations, enabling separation of physiological and non-physiological components; Lindeman-Merenda-Gold (LMG) relative importance decomposition further identified the dominant control on SIF variation. In parallel, three-band vegetation indices (TBIs) tailored to stress-specific spectral signatures were developed, and compact composite indices were formed by algebraic coupling of SIF and its derivatives (<span><math><mi>ΦF</mi><mo>-</mo><mi>p</mi></math></span>,<span><math><mspace></mspace><msub><mi>SIF</mi><mi>yield</mi></msub></math></span>, <span><math><msub><mi>SIF</mi><mi>total</mi></msub></math></span> and <span><math><msub><mi>SIF</mi><mi>mea</mi></msub></math></span>) with TBIs through addition, multiplication, and exponentiation; four regression families (linear, quadratic polynomial, logarithmic, and exponential) were evaluated for DI retrieval. With escalating disease severity, the contribution of <span><math><mi>ΦF</mi><mo>-</mo><mi>f</mi></math></span> increased monotonically from 51.501% to 89.444%, indicating that disease-driven physiological modulation is the principal source of SIF variability. Four TBIs, namely TBI-3 (554, 890, 710), TBI-7 (762, 866, 702), TBI-11 (758, 878, 542), and TBI-16 (918, 878, 542), exhibited strong correlations with the disease index, reaching 0.8678, 0.8674, −0.8423, and −0.8393, respectively. Composite indices combining the two parameter families further strengthened DI relationships; quadratic fits for <span><math><mi>ΦF</mi><mo>-</mo><mi>p</mi><mo>∗</mo><mo>∗</mo><mi>TBI</mi><mo>-</mo><mn>3</mn></math></span>, <span><math><mi>ΦF</mi><mo>-</mo><mi>p</mi><mo>∗</mo><mo>∗</mo><mi>TBI</mi><mo>-</mo><mn>7</mn></math></span>, <span><math><msub><mi>SIF</mi><mi>yield</mi></msub><mo>∗</mo><mo>∗</mo><mi>TBI</mi><mo>-</mo><mn>3</mn></math></span> produced R<sup>2</sup> minima of 0.800, 0.815, and 0.831 and maxima of 0.832, 0.836, and 0.841, with narrow spreads confirming robustness to data partitioning. Overall, these findings provide a mechanistic perspective on SIF variability in maize leaf spot and demonstrate the potential of integrating SIF-derived parameters with three-band vegetation indices for disease severity quantification under field conditions.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 2011-2033"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149347684","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
PVDT: A smart tool for rapid detection of pollen viability PVDT:快速检测花粉活力的智能工具
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-06-13 DOI: 10.1016/j.aiia.2026.05.011
He Li, Xuemei Du, Yilin Mao, Han Zhang, Hongtao Shi, Xinning Li, Juhong Zhang, Zihui Wang, Riliang Gu, Qun Sun
{"title":"PVDT: A smart tool for rapid detection of pollen viability","authors":"He Li,&nbsp;Xuemei Du,&nbsp;Yilin Mao,&nbsp;Han Zhang,&nbsp;Hongtao Shi,&nbsp;Xinning Li,&nbsp;Juhong Zhang,&nbsp;Zihui Wang,&nbsp;Riliang Gu,&nbsp;Qun Sun","doi":"10.1016/j.aiia.2026.05.011","DOIUrl":"10.1016/j.aiia.2026.05.011","url":null,"abstract":"<div><div>Pollen viability is a critical determinant of maize reproductive capacity and genetic potential, directly influencing pollination efficiency, seed set, and overall crop yield. While traditional detection methods are labor-intensive and time-consuming, rapid alternatives like flow cytometry remain prohibitively expensive, and existing machine vision approaches lack specialized tools. To address these challenges, the study introduces “AIpollen” software and a cost-effective Pollen Viability Detection Tool (PVDT) device. The software employs a dual-pathway detection system: (1) the YOLOv10-CBAM model, which achieves 90%–98% accuracy even under challenging conditions (e.g., variable lighting, high pollen density and low image quality), outperforming existing advanced models; and (2) the Pollen Microscopic Image Segmentation Algorithm (PMISA). When combined with a Residual Dual-Attention Gated Network (RDAGNet) model, PMISA mitigates performance degradation in high-density or low-quality images, boosting detection accuracy by approximately 7%. Field trials utilizing the PVDT evaluated pollen viability in response to heat stress for 350 maize inbred lines, supporting the exploration of loci associated with pollen viability through GWAS. Furthermore, Transfer learning enabled the detection of pollen viability across ten additional plant species (e.g., wheat, peas, alfalfa, potato, tobacco) with favorable detection performance.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 1595-1612"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149347781","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Unsupervised geometry-driven DSM terrain correction for Rice seedling segmentation in flooded Paddy fields 水稻田稻田秧苗分割的无监督几何驱动DSM地形校正
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-06-11 DOI: 10.1016/j.aiia.2026.04.014
He Zheng, Dongfang Li, Xiaomei Xu, Longzhe Quan, Yejun Zhu, Maohua Xiao
{"title":"Unsupervised geometry-driven DSM terrain correction for Rice seedling segmentation in flooded Paddy fields","authors":"He Zheng,&nbsp;Dongfang Li,&nbsp;Xiaomei Xu,&nbsp;Longzhe Quan,&nbsp;Yejun Zhu,&nbsp;Maohua Xiao","doi":"10.1016/j.aiia.2026.04.014","DOIUrl":"10.1016/j.aiia.2026.04.014","url":null,"abstract":"<div><div>Rice seedlings in flooded paddy fields are small with low ground cover. Specular reflections and weakened texture features substantially compromise their distinguishability in remote sensing imagery. Furthermore, digital surface models (DSMs) generated from unmanned aerial vehicle (UAV) imagery often contain systematic trend errors, which hinder reliable seedling segmentation. To address these challenges, we propose an unsupervised, geometry-driven method for segmenting rice seedlings from UAV imagery. This method formulates rice seedling segmentation as a geometry-driven seedling structure segmentation problem. An uncertainty-weighted DSM terrain-structure decomposition model is constructed for this purpose. By incorporating second-order total variation constraints, it corrects low-frequency terrain trends, thereby separating ground trends from sparse structures such as seedlings under unsupervised conditions. Moreover, the corrected geometric information is fused with morphological cues from RGB imagery to achieve stable segmentation of rice seedlings. This approach eliminates the need for ground control points or pixel-level manual annotations and effectively suppresses interference from water surface reflections and complex backgrounds. Experiments across multiple typical paddy field scenarios show that our method achieves seedling segmentation accuracy comparable to that of state-of-the-art fully supervised methods under complex conditions, attaining an mIoU of approximately 0.85. In addition, it exhibits stable structural consistency and robustness, validating its feasibility for large-scale paddy field monitoring and autonomous agricultural machinery navigation applications.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 1567-1581"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149347791","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
CTNet: A cognitive-inspired TriPercept CrossFusion self-distillation model for image classification in agricultural field CTNet:基于认知启发的农业领域图像分类TriPercept CrossFusion自蒸馏模型
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-05-30 DOI: 10.1016/j.aiia.2026.05.003
Bo Han, Heguang Sun, Wenqian Tang, Huanming Mai, Changsheng Li, Yi Zhang, Zhangjun Kou, Haofeng Qiu, Yubin Lan, Cunjia Liu, Xiaoling Deng
{"title":"CTNet: A cognitive-inspired TriPercept CrossFusion self-distillation model for image classification in agricultural field","authors":"Bo Han,&nbsp;Heguang Sun,&nbsp;Wenqian Tang,&nbsp;Huanming Mai,&nbsp;Changsheng Li,&nbsp;Yi Zhang,&nbsp;Zhangjun Kou,&nbsp;Haofeng Qiu,&nbsp;Yubin Lan,&nbsp;Cunjia Liu,&nbsp;Xiaoling Deng","doi":"10.1016/j.aiia.2026.05.003","DOIUrl":"10.1016/j.aiia.2026.05.003","url":null,"abstract":"<div><div>In agricultural artificial intelligence image classification tasks, existing models often exhibit redundant feature extraction and limited generalization capability. To address these practical demands and technical challenges, we propose a Cognitive-Inspired TriPercept CrossFusion Self-Distillation Network (CTNet). First, to counteract the adverse effects of environmental variations and motion blur inherent in agricultural image acquisition, we introduce a comprehensive data augmentation strategy specifically designed for agricultural imagery, thereby enhancing model robustness. Second, we present an efficient and lightweight TriPercept CrossFusion Network (TPCF), which concurrently captures multi-branch representations and emphasizes discriminative local features. A cross-branch fusion mechanism facilitates inter-branch feature interaction, enabling effective feature learning with minimal computational overhead. Third, we devise a novel end-to-end trainable self-distillation framework, termed Cognitive-Inspired Multi-Stage Self-Distillation (CISD). CISD operates without external pretrained teacher models; instead, it performs intra-model knowledge transfer across layers and employs a composite objective combining classification loss (cross-entropy), distillation loss (KL divergence), and hint loss (feature map alignment). Inspired by cognitive load theory, the framework dynamically adjusts the weighting of these loss components to achieve cognitive-load-aware multi-stage self-distillation, thereby improving generalization. We evaluated CTNet on a custom lychee disease dataset and six public benchmarks. On our augmented lychee dataset, CTNet achieves 99.03% accuracy. On three normal-collected datasets (PlantVillage, CottonWeedID15, and PotatoLeafDisease) the model attains accuracies of 99.74%, 91.81%, and 99.52%, respectively. On three controlled-collected datasets (PlantSeedling, RiceImage, and SoyBeanSeed) the corresponding accuracies are 97.45%, 99.97%, and 93.08%. CTNet consistently outperforms state-of-the-art classification methods across all benchmarks, demonstrating its effectiveness as a practical solution for agricultural image classification tasks.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 1457-1475"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149347859","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
PSB-diff: A Bernoulli diffusion segmentation framework based on rice seedling root point annotations for centimeter level row extraction PSB-diff:基于水稻幼苗根点注释的cm级行提取伯努利扩散分割框架
IF 16.1
Artificial Intelligence in Agriculture Pub Date : 2026-09-01 Epub Date: 2026-05-27 DOI: 10.1016/j.aiia.2026.05.007
Yongchao Shan, Weiqiang Fu, Yanxin Yin, Anqi Zhang, Hao Wang, Yue Zhao, Zhijun Meng, Chunjiang Zhao
{"title":"PSB-diff: A Bernoulli diffusion segmentation framework based on rice seedling root point annotations for centimeter level row extraction","authors":"Yongchao Shan,&nbsp;Weiqiang Fu,&nbsp;Yanxin Yin,&nbsp;Anqi Zhang,&nbsp;Hao Wang,&nbsp;Yue Zhao,&nbsp;Zhijun Meng,&nbsp;Chunjiang Zhao","doi":"10.1016/j.aiia.2026.05.007","DOIUrl":"10.1016/j.aiia.2026.05.007","url":null,"abstract":"<div><div>Precise extraction of rice seedling rows from field-scale unmanned aerial vehicle (UAV) orthomosaics is crucial for precision agriculture. Existing methods typically rely on pixel-wise semantic segmentation with full plant-contour annotations to localize seedlings and extract row centerlines; however, such approaches require costly full-contour labeling and are often sensitive to complex paddy-field conditions, limiting practical deployment. To address these limitations, we propose PSB-Diff, a point-supervised Bernoulli diffusion segmentation framework driven by rice root-point annotations. A discriminative segmenter first generates coarse prior masks, which are progressively refined via prior-guided Bernoulli diffusion, confidence-aware denoising, and local position correction to improve root localization. Seedling centroids inferred from the refined masks are then grouped into continuous crop-row centerlines through principal-direction gridding, intra-grid adaptive between-row grouping, and crop-row cross-grid connection. Experiments on UAV orthomosaics from three rice-growing regions show that, when paired with a Transformer-based segmentation prior (SegFormer), PSB-Diff attains F1@8 = 91.9% (P@8 = 94.5%, R@8 = 89.5%) under fully annotated training and F1@8 = 89.8% (P@8 = 94.2%, R@8 = 85.8%) when pseudo-labels replace 75% of human labels, indicating strong robustness to label noise under limited supervision. Crop-row evaluation yields median and mean centerline distance errors of 1.80 cm and 1.95 cm, satisfying centimeter-level accuracy. Overall, PSB-Diff supports accurate seedling localization and crop-row extraction at low annotation cost, with practical relevance for precision-agriculture applications.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 3","pages":"Pages 1519-1535"},"PeriodicalIF":16.1,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149347863","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
相关产品
×
本文献相关产品
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信
小红书