Smart Leaf Identification and Analysis using Artificial Intelligence
构建叶片分类框架,比较多种机器学习与深度学习方法识别36种叶片。
构建叶片分类框架,比较多种机器学习与深度学习方法识别36种叶片。
系统综述253个农业计算机视觉数据集,提出多维分类体系并指出局限。
该综述基于PRISMA流程从2308条记录中筛出253个数据集出版物,提出按任务、语义与设计特征划分的统一分类法,可用于数据集选型;但结论限于Scientific Data与Data in Brief两刊,覆盖面有边界。
提出CNN框架对水果和作物叶片进行五级品质分级与缺陷识别。
该预印本用单一CNN同时判别成熟度与病虫害、腐烂等缺陷,五分类测试准确率94.9%,单图约21ms;但数据规模与作物种类未在摘要披露,且未经同行评审,结论宜作参考而非定论。
Weight-posture joint modeling for body weight estimation of unconstrained pigs。Smart Agricultural Technology
该研究针对无约束养殖场景,将体重与姿态联合建模以提升猪只估重精度,适用于非接触式养殖监测,但实际精度与数据规模需看原文验证。
AI-enabled vision-based perception models are increasingly used in precision agriculture for tasks such as fruit detection, trunk detection, and autonomous navigation. As these models move into field deployment, their errors no longer stay confined to the image: they propagate into the counts, maps, and yield estimates that downstream automation depends on, making adversarial robustness a security concern. Yet this robustness remains underexplored in orchard settings. Using the field-deployed Agrosense v2 platform as a case study, we evaluate tree-trunk and fruit detection under Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) attacks across a range of perturbation strengths. The results are summarized over five runs as mean ± sample standard deviation (SD). Both detectors degrade sharply: under PGD at ϵ = 8 , trunk detection mAP@0.5 falls from 99.36% to (13.84 ± 2.95)%, while fruit detection mAP@0.5 falls to (0.54 ± 0.16)%. These errors also propagate downstream, with correctly preserved detections dropping from 274 to 15.6 ± 3.6 for trunks and from 4058 to 9.8 ± 1.3 for fruit relative to the clean pipeline. We assess three defense families. Fine-tuning improves performance on transferred confidence-suppression attacks, but stronger native-loss and defense-aware attacks expose substantial remaining vulnerability. Multi-budget training improves clean fruit detection accuracy at the cost of robustness at higher attack strength. We have considered transfer-based attacks, image corruption, and performance/memory overhead to evaluate the effectiveness of these defenses. These results show that strong clean-image accuracy does not guarantee reliable behavior under attack, and robustness must be evaluated at the application level, not via detector metrics alone.
Non-Contact Weight Estimation of Piglets Using RGB Images and Explicit Geometric Priors via Cross-Attention Fusion。Smart Agricultural Technology
该研究用RGB图像结合显式几何先验与交叉注意力融合估计仔猪体重,为无接触称重提供新思路;但摘要未给出样本规模与误差指标,适用性需查原文验证。
Estimating swath width is essential for improving fertilizer distribution uniformity in precision fertilizer application. This study presents a rapid vision-based framework for swath-width prediction using a hybrid CNN–Transformer architecture trained on synthetic video sequences generated from a physics-informed simulation of fertilizer particle motion incorporating gravity and aerodynamic drag. The proposed method provides an alternative to tray-based calibration, ground-deposition analysis, and individual particle tracking, which are labor-intensive or unreliable under high particle overlap. The dataset included 297 synthetic videos sampled into 25 sequential frames, consisting of 162 baseline scenarios with variations in disc diameter, disc speed, and disc height, and 135 additional scenarios incorporating wind speed and wind direction. Spatial features were extracted using a ResNet-18 backbone, combined with temporal positional encoding, and processed by a Transformer encoder. Model performance was evaluated using held-out synthetic samples, independent EDEM simulations, laboratory collector-bin measurements, and real image sequences captured under controlled conditions. On held-out synthetic test data, the generalized model achieved an MAE of 0.529 m, RMSE of 0.695 m, R 2 of 0.996, MAPE of 8.06%, bias of − 0.142 m, and MaxAE of 1.51 m. Compared with EDEM simulations, it achieved an MAE of 0.535 m, RMSE of 0.594 m, R 2 of 0.954, MAPE of 10.61%, bias of − 0.40 m, and MaxAE of 0.99 m. Laboratory validation using real image sequences also showed close agreement between baseline-model predictions and measured swath widths. These results demonstrate the feasibility of the approach, although field-scale validation under terrain slope, dust, vegetation occlusion, and variable illumination is still required.
提出融合关键点检测与主动视角调整的莲蓬采摘机器人方法,并在实验室受控条件下验证。
提出边界感知与结构引导的由粗到细语义分割方法,用于绿色果园果实分割。
提出基于骨架拓扑约束与时序跟踪的多视角投影方法,估计水稻剑叶倾角。
提出YOLO-PLS模型,在冠层尺度实时检测超小叶斑病斑并实现边缘部署。
开发便携式YOLOv5鱼苗实时检测系统,结合浊度传感校正,用于废弃锡矿塘浑浊水体中的罗非鱼苗计数。
开发了PhenoDesk数字表型平台,实现叶片图像自动分割、形态测量与可重复分析。
研制主动分离近距成像装置并构建数据集,提出YOLO-PFL模型识别水稻秧苗叶龄。
提出AeroFlock Net,实现斜视航拍影像中放牧羊群的跨视角自适应检测与计数。
提出双编码器RGB-D模型,实时定位奶牛及背部解剖区域。
提出混合局部-全局视觉Transformer模型ViT-LGFNet用于植物病害检测。
用YOLOv8n自动量化断奶仔猪栏级采食器占用率,并比较两种栏型与性别差异。
提出基于计算机视觉的种蛋三级受精分类框架,区分高受精、受精与未受精。
将计算机视觉果实检测与多变量基因组选择结合,提升蓝莓产量预测精度。