https://www.alice.cnptia.embrapa.br/alice/handle/item/197 Artigo em anais de congresso (CNPTIA) Fri, 19 Jun 2026 23:59:34 GMT 2026-06-19T23:59:34Z https://www.alice.cnptia.embrapa.br/alice/handle/doc/1187153 Título: Cattle weight estimation from dense point clouds. Autoria: CASTANHEIRO, L. F.; TETILA, E. C.; FURUYA, D. E. G.; SILVA, J. P. da; BARBEDO, J. G. A.; ROMANI, L. A. S.; BOLFE, E. L. Conteúdo: Cattle weight is essential for decision-making in precision livestock farming, directly supporting nutrition management, animal welfare, and production efficiency. Existing methods rely on close-range measurements or manual intervention, limiting scalability. This work proposes an workflow for cattle weight estimation based on point clouds derived from aerial images. RGB images acquired at low altitude were processed using Structure from Motion (SfM) techniques to generate dense point clouds. Individual animals were automatically segmented from the reconstructed 3D scene, and voxel-based volumetric features were extracted for each animal. Body weight was then estimated through linear regression models calibrated with ground truth measurements obtained from individual weighing. The proposed approach was evaluated on Nellore cattle in a feedlot environment and achieved a root mean square error (RMSE) of 8.35 kg, corresponding to an average relative error of approximately 2.29%. The results highlight the potential of UAV-based photogrammetry as a cost-effective decision support tool for digital and sustainable livestock management. Thu, 01 Jan 2026 00:00:00 GMT https://www.alice.cnptia.embrapa.br/alice/handle/doc/1187153 2026-01-01T00:00:00Z https://www.alice.cnptia.embrapa.br/alice/handle/doc/1185950 Título: SEEmear: a system for large scale geo-referenced stereo imaging of orchards. Autoria: SANTOS, T. T.; KOENIGKAN, L. V.; GEBLER, L. Conteúdo: Precision agriculture applications at the plant level require high-resolution, detailed imaging that cannot be adequately provided by satellite-based remote sensing due to insufficient resolution and suboptimal viewing angles. While Unmanned Aerial Vehicles (UAVs) offer improved capabilities, their downward-facing perspective and navigation challenges in orchards limit their effectiveness for fruit monitoring and anomalies detection. To address these limitations, we present SEEmear, a novel ground-based proximal sensing system for large-scale geo-referenced stereo imaging of orchard environments. The system integrates high-performance embedded computing, wide-angle global shutter RGB-D cameras, and precision RTK GNSS positioning, enabling simultaneous imaging of both sides of orchard rows from close proximity. SEEmear's 110° field-of-view cameras capture entire tree structures even at distances of 1 meter, while global shutter sensors eliminate motion artifacts essential for moving platforms. We tested the system in apple orchards, collecting comprehensive geo-referenced RGB-D imagery across 1.50 ha in approximately 40 minutes per session. The resulting dataset supports advanced applications including depth estimation, 3D reconstruction, background filtering, and object segmentation. The adaptable platform integrates with various ground vehicles and provides substantial data storage and processing capabilities needed for hardware-accelerated AI algorithms. SEEmear addresses the critical need for high-quality proximal sensing data in precision agriculture research, supporting applications in fruit detection, tracking, yield mapping, autonomous navigation, and field robotics. Wed, 01 Jan 2025 00:00:00 GMT https://www.alice.cnptia.embrapa.br/alice/handle/doc/1185950 2025-01-01T00:00:00Z https://www.alice.cnptia.embrapa.br/alice/handle/doc/1185938 Título: UAV-based automation: a case study of coffee crop input application in Caconde, Brazil. Autoria: SILVA, T. L. da; ROMANI, L. A. S.; MARTHA JUNIOR, G. B.; GREGO, C. R.; LUCHIARI JUNIOR, A.; MASSRUHÁ, S. M. F. S. Conteúdo: Precision agriculture optimizes agricultural practices by applying inputs at the right time and place, enhancing productivity and sustainability. This study compared the costs, time requirements, and water efficiency of using an unmanned aerial vehicle (UAV) versus the traditional manual backpack sprayer method on a 5-hectare coffee farm in Caconde, São Paulo, Brazil. Results showed that UAV application reduced operational costs by 21.8% (R$750 vs. R$960), water consumption by 96% (50 liters vs. 1,250 liters), and application time by 75% (two hours vs. eight hours). Sensitivity analysis confirmed the cost-effectiveness of UAV method under varying labor costs. This study underscores the economic and environmental advantages of UAV spraying and highlights its potential to address labor shortages and promote sustainable precision farming. Wed, 01 Jan 2025 00:00:00 GMT https://www.alice.cnptia.embrapa.br/alice/handle/doc/1185938 2025-01-01T00:00:00Z https://www.alice.cnptia.embrapa.br/alice/handle/doc/1185942 Título: Semear MAppleT FW: a dataset for apple detection and tracking in orchards under fruiting wall training system. Autoria: SANTOS, T. T.; FARIA, L. N. de; GEBLER, L. Conteúdo: Computer vision techniques for fruit detection and tracking are crucial for agricultural automation, yet most current datasets lack temporally consistent annotations needed for reliable tracking. Here we present Semear MAppleT FW, a dataset for apple detection and tracking in modern fruiting wall systems. The dataset comprises six video sequences of 100 frames each, captured by two RGB-D stereo cameras mounted on a tractor traversing orchard rows. Unlike previous datasets, Semear MAppleT FW features wide-angle images capturing entire tree lengths, ensuring complete canopy visibility within the field of view. To date, we provide over 53,000 bounding box annotations for 1,267 unique apple instances with temporal consistency across frames, stereo image pairs with known baseline calibration, and 3D reconstruction data. Our annotation method leverages structure-from-motion to estimate fruit positions in 3D space, enabling accurate tracking even when fruits are occluded by branches, leaves, or other fruits. The dataset includes visibility flags for each annotation, distinguishing between visible and occluded fruits. This approach maintains spatial consistency of annotations across frames while significantly reducing manual annotation workload. Semear MAppleT FW provides a valuable resource for developing artificial intelligence systems for automated yield estimation, fruit growth monitoring, and robotic harvesting in commercial orchards. Wed, 01 Jan 2025 00:00:00 GMT https://www.alice.cnptia.embrapa.br/alice/handle/doc/1185942 2025-01-01T00:00:00Z