Artificial Intelligence in Agriculture最新文献

{"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}

{"title":"Spatial pose estimation of apples for robotic harvesting","authors":"Chunlin Chen,&nbsp;Furui Zhang,&nbsp;Dezhao Huang,&nbsp;Zhuoying Song,&nbsp;Qingyan Yang,&nbsp;Fuzeng Yang,&nbsp;Zheng Wang","doi":"10.1016/j.aiia.2026.01.003","DOIUrl":"10.1016/j.aiia.2026.01.003","url":null,"abstract":"<div><div>In the apple picking process, the existing robots are unable to accurately estimate the pose of apples. This often leads to damage to the fruits, branches, and even the robot fingers during grasping, while also significantly reducing the efficiency and success rate of apple picking. Hence this paper proposes a real-time 3D attitude angle estimation algorithm for apples in orchards and has independently developed a set of apple posture angle verification system. Finally, through field trials in the orchard, the effectiveness and accuracy of the algorithm are verified. Firstly, based on the fruit segmentation mask map of YOLOv8 combined with the moment of inertia algorithm, the pose angle (<span><math><mn>0</mn><mo>,</mo><mi>θ</mi></math></span>) and its maximum transverse diameter of the invisible calyx apple are estimated. When the model recognizes the apple calyx, the position of the core is estimated based on the surface center coordinates and the maximum transverse diameter of the apple, and then the calyx-core vector is constructed. Finally, through spatial geometry, the vectors are decomposed into the <span><math><msup><mi>X</mi><mrow><mo>'</mo><mo>'</mo></mrow></msup><msup><mi>O</mi><mrow><mo>'</mo><mo>'</mo></mrow></msup><msup><mi>Y</mi><mrow><mo>'</mo><mo>'</mo></mrow></msup></math></span> plane and the <span><math><msup><mi>Y</mi><mrow><mo>'</mo><mo>'</mo></mrow></msup><msup><mi>O</mi><mrow><mo>'</mo><mo>'</mo></mrow></msup><msup><mi>Z</mi><mrow><mo>'</mo><mo>'</mo></mrow></msup></math></span> plane respectively, and then the attitude angles (<span><math><mi>α</mi><mo>,</mo><mi>β</mi></math></span>) are calculated. The field test results of the orchard show that the accuracy rates of this algorithm in detecting apples and calyxes are 0.953 and 0.703 respectively, and the accuracy rate of fruit segmentation is 0.953. Through the calibration error test experiment, the spatial positioning error of the robotic arm is 12.1 mm. After multi-pose and multi-distance tests, the absolute error of the algorithm in estimating the maximum transverse diameter of the apple is 3.9 mm. Field experiments in orchards were conducted through the self-developed attitude angle verification system. The average absolute error of the attitude angle of apples without identified calyxes was 16.3°. Among the apple pose angles that can identify the calyx, the average absolute error is 17.0° and the average absolute error is 18.1°. The Apple pose estimation algorithm proposed in this paper provides algorithmic support for the subsequent pose capture of apple picking robots.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 2","pages":"Pages 1227-1242"},"PeriodicalIF":16.1,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148625306","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}

{"title":"EDSC-HRAFNet: An apple tree branch semantic segmentation model for harvesting robots under complex orchard conditions","authors":"Zhengyang Liu,&nbsp;Qingchun Feng,&nbsp;Chengjin Qin,&nbsp;Yangtian Lin,&nbsp;Pengcheng Xia,&nbsp;Haodi Wang,&nbsp;Liang Gong,&nbsp;Chengliang Liu","doi":"10.1016/j.aiia.2026.01.005","DOIUrl":"10.1016/j.aiia.2026.01.005","url":null,"abstract":"<div><div>Accurate semantic understanding of tree branches is critical for orchard harvesting robots in automated fruit harvesting and pruning. Existing methods suffer from low detection accuracy and limited adaptability in complex orchard environments. This paper proposes a novel apple tree branch semantic segmentation model (EDSC-HRAFNet) for orchard harvesting robots under complex orchards conditions. The presented Enhanced Dynamic Snake Convolution module (EDSC_unit) is integrated into High-Resolution Network (HRNet) backbone to extract topological features such as branching points and bifurcations. Then, the HeteroFPN module is designed as the Neck structure, and performs semantic-position information cyclic interaction on the multi-level output features of Backbone in a dual-path collaborative framework, obtaining multi-level features with stronger comprehensive representation capabilities. And the Parallel-M4 Decode module is designed for the network head, performing parallel processing based on the characteristics of features at different levels. This framework could concatenate the features to generate geometrically precise segmentation masks. Finally, we constructed a dataset of in-situ apple trees under diverse real-world conditions to verify the performance and superiority of EDSC-HRAFNet. EDSC-HRAFNet demonstrates state-of-the-art segmentation performance across eight challenging orchard scenarios and exhibits robust generalization. Experiments show that proposed model achieves precision, recall, dice, IoU, MIoU, and MPA of 91.50%, 91.71%, 91.60%, 84.51%, 91.72%, and 95.58%, respectively. These improvements are 6.98%, 10.09%, 9.21%, 13.5%, 7.25%, and 5.24% compared to HRNet. Compared with existing models including Pspnet, Deeplabv3+ series and Unet series, the precision, recall, IoU is improved by 13.92% to 29.61%, 23.59% to 42.38% and 28.79% to 46.67% respectively. EDSC-HRAFNet's superior branch segmentation capability in challenging orchard environments provides a practical foundation for robotic automation in agricultural orchards.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 2","pages":"Pages 1256-1280"},"PeriodicalIF":16.1,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148625360","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}

{"title":"Soil knowledge-guided multi-task transformer model for predicting soil properties and crop traits with early warning alerts","authors":"Guangyao Sun,&nbsp;Wencai Zhang,&nbsp;Yuanhao Lou,&nbsp;Tonglin Liao,&nbsp;Chen Chen,&nbsp;Lisuo Ren,&nbsp;Zengqiang Duan,&nbsp;Chunli Lv,&nbsp;Xiangbin Kong,&nbsp;Wenjun Ji,&nbsp;Yuntao Ma","doi":"10.1016/j.aiia.2026.01.001","DOIUrl":"10.1016/j.aiia.2026.01.001","url":null,"abstract":"<div><div>Soil properties are vital for soil health, fertility, and crop productivity, yet traditional methods like the Partial Least Squares Regression (PLSR) predict individual properties without capturing their interactions or impacts on crop growth. This study presents the Soil Knowledge-Guided Multi-task Transformer (KGMT) model, which integrates deep learning and multi-task learning to simultaneously predict multiple soil properties and crop traits using UAV hyperspectral imagery, eliminating complex manual feature extraction. By embedding soil knowledge, KGMT improves prediction accuracy and generalization over PLSR and baseline multi-task models. Data from two agricultural plots in Bayannur City, Inner Mongolia, China, demonstrated KGMT's superior performance in predicting six soil properties (alkalinity, salinity, organic matter, potassium, sodium, calcium) and two crop traits (leaf area index, aboveground biomass), achieving up to 20% lower normalized root mean square error. For instance, KGMT yielded higher R² values for alkalinity (0.57) and organic matter (0.57) compared to PLSR (0.46 and 0.40, respectively). The model also enhanced cross-region generalization. Additionally, a soil early warning framework was developed, linking pre-planting soil conditions with crop traits to detect potential growth limitations early, enabling proactive management. This approach integrates spectral data with domain knowledge, offering a robust solution for soil and crop management. It contributes to precision agriculture by optimizing soil conditions and promoting sustainable land use practices, highlighting the value of advanced deep learning in agricultural applications.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 2","pages":"Pages 1197-1211"},"PeriodicalIF":16.1,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148625362","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}

{"title":"AgriGaussian: A low-cost 3D reconstruction method for high-fidelity plant architecture analysis","authors":"Qilin An,&nbsp;Guangrui Li,&nbsp;Xue Deng,&nbsp;Yongzhi Cui,&nbsp;Quanfeng Guo,&nbsp;Yangchun Liu,&nbsp;Liming Zhou,&nbsp;Liguo Wei,&nbsp;Bo Zhao","doi":"10.1016/j.aiia.2026.01.006","DOIUrl":"10.1016/j.aiia.2026.01.006","url":null,"abstract":"<div><div>High-quality, real-time three-dimensional (3D) reconstruction and phenotyping of field crops are essential for advancing digital farm management. However, challenges such as diverse crop postures and fluctuating illumination in open-field environments significantly hinder reconstruction accuracy. To address these issues, this study proposes AgriGaussian, a novel 3D phenotyping method based on 3D Gaussian Splatting (3DGS), designed to enable low-cost, high-fidelity reconstruction and phenotypic analysis of agricultural fields, with chili pepper plants as an example. Specifically, a dataset encompassing three distinct growth stages of chili pepper plants was collected and preprocessed using a phenotyping robot (PR) equipped with oblique-view imaging. The AgriGaussian method integrates three core components: (1) a Depth-Supervised Strategy (DSS) for adaptively enhancing phenotypic features based on gradient variations; (2) a Crop Appearance Extraction (CAE) module that suppresses the effects of illumination and environmental interference while preserving fine phenotypic details; and (3) a Multi-scale Chunking Training (MCT) strategy that enables scalable reconstruction of large agricultural scenes. Experimental results demonstrate that AgriGaussian achieves high-fidelity reconstruction, with peak signal-to-noise ratio (PSNR) of 26.85 dB, structural similarity index (SSIM) of 0.86, and learned perceptual image patch similarity (LPIPS) of 0.24—surpassing existing baseline algorithms. Compared to manual measurements, the reconstructed canopy height and volume exhibit strong agreement, with coefficients of determination (R²) reaching 0.90 and 0.86, respectively, and average absolute percentage errors as low as 5.1%. Furthermore, the method accurately reconstructs fine-grained features such as foliar lesions and surface textures. In summary, AgriGaussian enables large-scale, high-fidelity reconstruction of field crop phenotypes and provides a promising foundation for advancing 3D phenotyping and digital twin technologies in precision agriculture.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 2","pages":"Pages 1281-1295"},"PeriodicalIF":16.1,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148625307","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}

{"title":"Decoupling canopy structure effects from vegetation indices for robust assessment of potato plant nitrogen content across growth stages","authors":"Yiguang Fan,&nbsp;Haikuan Feng,&nbsp;Riqiang Chen,&nbsp;Yang Liu,&nbsp;Yang Meng,&nbsp;Jiejie Fan,&nbsp;Mingbo Bian,&nbsp;Yanpeng Ma,&nbsp;Guijun Yang,&nbsp;Chunjiang Zhao","doi":"10.1016/j.aiia.2026.02.001","DOIUrl":"10.1016/j.aiia.2026.02.001","url":null,"abstract":"<div><div>Accurate monitoring of plant nitrogen concentration (PNC) is essential for precision nitrogen (N) fertilization management at the farm scale. While numerous vegetation indices (VIs) have been proposed for PNC monitoring, their applicability and accuracy often vary across growth stages. The impact of canopy structural variations across phenological stages on the performance of VIs in estimating crop N status has been confirmed, but the underlying mechanisms and effective solutions remain insufficiently understood. This study systematically analyzed how growth stage-induced canopy structural variations affect potato PNC estimation using VIs. Furthermore, a novel index named the canopy structure adjustment vegetation index (CSAI) was developed to mitigate the interference of growth stages in PNC estimation. During the 2022 and 2023 growing seasons, a total of 432 PNC measurements and corresponding unmanned aerial vehicle (UAV) multispectral images were collected from potato experimental fields in Northeast China. Experimental plots with N and potassium treatments were laid out to rigorously evaluate the performance of CSAIs in estimating the PNC of potato. Our findings revealed that fractional vegetation cover (FVC) was the primary driver of VIs variations across different potato growth stages, thus rendering PNC estimation by VIs strongly dependent on the relationship between FVC and PNC. The newly developed CSAI (CSAI = VI/FVC) can better adapt to the dynamic growth patterns of potatoes, substantially mitigating growth stage effects on PNC estimation. Validation using an independent seasonal dataset indicated that CSAIs derived from the optimized soil-adjusted vegetation index (OSAVI) and normalized difference red edge index (NDRE) yielded higher <em>R</em>² values of 0.88 and 0.80, with lower RRMSE values of 13.88% and 16.65%, respectively. Using potato as an example, this study investigated the mechanisms underlying how growth stages influence the performance of UAV-based VIs in estimating PNC and introduced effective solutions. Future research should further validate the CSAI framework across broader environmental conditions and crop species. These findings provide a promising foundation for informed decision-making in crop N management.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 2","pages":"Pages 1332-1346"},"PeriodicalIF":16.1,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148625377","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}

{"title":"Zero-shot crop segmentation via 3D depth-aware vision pipeline using unsupervised clustering","authors":"Hassan Afzaal ,&nbsp;Aitazaz A. Farooque ,&nbsp;Gurjit S. Randhawa ,&nbsp;Arnold W. Schumann ,&nbsp;Nicholas Krouglicof ,&nbsp;Yuvraj S. Gill ,&nbsp;Qamar U. Zaman","doi":"10.1016/j.aiia.2026.05.001","DOIUrl":"10.1016/j.aiia.2026.05.001","url":null,"abstract":"<div><div>Pixel-level crop segmentation is essential for precision agriculture yet remains severely constrained by the prohibitive cost and labor intensity of generating large-scale annotated datasets, a critical bottleneck limiting autonomous crop monitoring adoption across agricultural systems. This study presents a paradigm-shifting, zero-shot segmentation framework that eliminates pixel-level annotation requirements entirely by leveraging monocular depth estimation models (Depth Anything, Depth Pro, DPT Large) combined with unsupervised clustering algorithms to segment crops based on 3D structural geometry. Evaluated on a comprehensive agricultural dataset of 19,940 images spanning potatoes, kidney beans, and soybeans across distinct agro-climatic regions of Prince Edward Island and Ontario, Canada, the proposed depth-aware pipeline substantially outperformed classical segmentation baselines and competed favorably with emerging foundation model approaches. The method achieved a mean F1-score of 0.875 and Intersection over Union (IoU) of 0.803, representing 93% to 164% improvement over traditional techniques (Otsu thresholding, Watershed, K-means), while maintaining superior consistency across crop types (standard deviation 0.057 versus 0.127 for vision-language models). Notably, Depth Anything combined with unsupervised Mini-Batch K-means clustering achieved this performance with an inference latency of 400 ms per image, making it practical for real-time field deployment. The framework's robustness stems from its utilization of geometric priors and inherent depth discontinuities, which differentiate crop vegetation from soil backgrounds under complex conditions where traditional color-based and even zero-shot vision-language models struggle. By circumventing the data scarcity trap that has plagued supervised deep learning in agricultural applications, this work establishes 3D structural reasoning as a scalable, highly accurate, and annotation-free solution for autonomous crop monitoring, directly advancing the accessibility of AI-driven precision agriculture for small and large-scale farming operations worldwide.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 2","pages":"Pages 1181-1196"},"PeriodicalIF":12.4,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148184607","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"LUNet: Language-infused UNet for precise Camellia oleifera pests and diseases segmentation in complex agricultural environment","authors":"Ziyang Shi,&nbsp;Hao Zhou,&nbsp;Lin Li,&nbsp;Shaofeng Peng,&nbsp;Ruifeng Liu,&nbsp;Fangying Wan,&nbsp;Jiangzhang Zhu,&nbsp;Ruopeng Zhang,&nbsp;Jiale Chen,&nbsp;Leheng Li","doi":"10.1016/j.aiia.2026.02.003","DOIUrl":"10.1016/j.aiia.2026.02.003","url":null,"abstract":"<div><div>With the rapid development of deep learning, precise segmentation of Camellia oleifera pests and diseases has become a pivotal technical bottleneck in crop health monitoring and intelligent plant protection decision-making. Although deep learning methods have been widely applied to plant disease detection, three major challenges remain prominent in agricultural scenarios: (1) the insufficient capability of conventional single-modal models to capture subtle pathological features; (2) the misclassification of disease symptoms due to complex environmental interference; and (3) suboptimal performance resulting from inadequate integration of global contextual information. To tackle these issues, this study introduces a novel multi-modal segmentation network, LUNet, as well as an intelligent diagnosis system. First, a vision-semantic fusion framework is developed, leveraging a cross-modal transformer to align visual features, thereby effectively integrating semantic information across various disease types. Second, an innovative Positional Mixed Attention (PMA) block is introduced to enhance the model's ability to locate disease spot boundaries under strong lighting interference. Third, a Multi-scale Feature Fusion (MFF) block is designed to improve the feature discrimination in scenarios with overlapping leaves. Furthermore, to tackle data scarcity, a specialized dataset encompassing 7 typical Camellia oleifera diseases (comprising 1400 high-resolution field images of diseased Camellia oleifera) is collaboratively developed with agricultural experts. Experimental results demonstrate that LUNet outperforms baseline models, improving the Dice coefficient and mIoU by 4.19% and 5.12%, respectively, and achieving a 15.52% higher segmentation mIoU than DeepLabV3+ in complex backgrounds. Building on these results, an intelligent diagnosis system for camellia oleifera pests and diseases is developed by integrating GPT-4 image understanding and DeepSeek-R1 deep reasoning modules. This system establishes a three-stage technical pipeline, including disease segmentation, pathological analysis, and control decision-making, thereby providing a practical and scalable solution for smart agriculture. Our models and test dataset are available at: <span><span>https://github.com/shziyang/LUNet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 2","pages":"Pages 1363-1383"},"PeriodicalIF":16.1,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148625380","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}

{"title":"AI-based respiratory rate estimation in group-housed pigs infected with Mycoplasma hyopneumoniae under occlusion conditions","authors":"Jin He,&nbsp;Qifeng Li,&nbsp;Ligen Yu,&nbsp;Taiyong Yu,&nbsp;Zhiyu Ren,&nbsp;Feng Qiu,&nbsp;Xiaoli Ding,&nbsp;Gan Yang,&nbsp;Chunjiang Zhao,&nbsp;Yanrong Zhuang","doi":"10.1016/j.aiia.2026.01.002","DOIUrl":"10.1016/j.aiia.2026.01.002","url":null,"abstract":"<div><div>Respiratory rate (RR) is a critical physiological indicator for assessing the health and welfare of pigs. Although numerous studies have focused on RR monitoring, most have overlooked the issue of mutual occlusion among group-housed pigs, and few have examined RR changes under <em>Mycoplasma hyopneumoniae</em> (Mhp) infection. This study addressed these gaps by proposing a computer vision-based method for estimating the RR of group-housed pigs infected with Mhp. An object detection model was used to identify individual pigs, and a region of interest (ROI) was automatically selected, within which RR was estimated based on periodic fluctuations in R, G, and B channel intensities. To reduce interference from neighboring pigs, an occlusion-handling mechanism was introduced. The results showed high consistency between estimated and actual RR, with a mean absolute error (MAE) of 1.38 breaths per minute (bpm), root mean square error (RMSE) of 2.16 bpm, mean absolute percentage error (MAPE) of 5.92%, and a correlation coefficient of 0.99. Statistical analysis revealed no significant difference in MAPE between occluded and unoccluded pigs, confirming the effectiveness of the occlusion mechanism. Furthermore, the model exhibited robust estimation accuracy under low-illumination nighttime conditions (MAE = 1.52 bpm, RMSE = 2.7 bpm, <em>R</em> = 0.98), further validating its applicability to long-term continuous monitoring. In addition, the observed RR patterns showed a clear association with the progression of Mhp infection. Overall, the proposed method offers a novel and effective approach for RR estimation and early warning of Mhp infection. The method used in this study has been made publicly available on GitHub (<span><span>https://github.com/candicejinhe/pig-breath</span><svg><path></path></svg></span>).</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 2","pages":"Pages 1212-1226"},"PeriodicalIF":16.1,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148624296","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}

{"title":"Integrative hormone-metabolite tracking reveals ultra-early immune responses to anthracnose in tea cultivars","authors":"Yan Hu,&nbsp;Zhenxiong Huang,&nbsp;Mostafa Gouda,&nbsp;Yiqiang Zhang,&nbsp;Xuecheng Zhang,&nbsp;Fengle Zhu,&nbsp;Sitan Ye,&nbsp;Liang He,&nbsp;Xiaoli Li,&nbsp;Yong He","doi":"10.1016/j.aiia.2026.01.004","DOIUrl":"10.1016/j.aiia.2026.01.004","url":null,"abstract":"<div><div>Plant fungal pathogens, particularly anthracnose, pose a significant threat to tea production by adversely affecting both yield and quality. Detecting these pathogens has been challenging due to the subtlety of their symptoms, making early identification critical for mitigating damage. The purpose of this study was to develop a novel, rapid, non-destructive, and in situ method for the ultra-early detection of tea anthracnose. The study employed hyperspectral imaging (HSI) in combination with the HybridSN-CBAM deep learning algorithm to detect tea anthracnose. A hormone–metabolite analysis was conducted to explore spectral response mechanisms, and a PLSR-based quantitative model was developed to correlate spectral data with key metabolic markers. Additionally, transmission electron microscopy (TEM) was used to examine cellular responses in resistant tea varieties. The results revealed that the HybridSN-CBAM method could detect anthracnose at least 12 h earlier than traditional methods. Metabolomics analysis identified key markers, including hormones, differential metabolites, and physiological indicators. A PLSR model demonstrated high predictive accuracy (Rp² &gt; 0.8) for these metabolites. TEM imaging showed that resistant tea varieties exhibited better structural preservation and enhanced defense mechanisms. This study demonstrates the potential of using HSI and deep learning for ultra-early, non-destructive plant disease detection. The findings contribute to precision agriculture by enabling real-time monitoring of plant health and promoting sustainable agricultural practices.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 2","pages":"Pages 1243-1255"},"PeriodicalIF":16.1,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148625375","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}