María Barrera-Báez
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
María Barrera-Báez has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Luis Sánchez-Fernández
Área de Ingeniería Agroforestal, Departamento de Ingeniería Aeroespacial y Mecánica de Fluidos, Universidad de Sevilla, 41005 Sevilla, Spain
2 shared publications - Manuel Pérez-Ruiz
Área de Ingeniería Agroforestal, Departamento de Ingeniería Aeroespacial y Mecánica de Fluidos, Universidad de Sevilla, 41005 Sevilla, Spain
2 shared publications A. Nizzoli
1 shared publication- Alessia Nizzoli
Área de Ingeniería Agroforestal, Departamento de Ingeniería Aeroespacial y Mecánica de Fluidos, Universidad de Sevilla, 41005 Sevilla, Spain
1 shared publication O.E Apolo-Apolo
1 shared publication- Orly Enrique Apolo-Apolo
Department of Environment, Faculty of Bioscience Engineering, Ghent University, 9000 Ghent, Belgium
1 shared publication
Latest Publications
- Open Access
- 11 August 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (16) , 7982
Integration of AI-Based Weed Detection and Robotic Actuation for Site-Specific Under-Canopy Spraying in Woody Crops
Weed management in woody perennial crops relies mainly on broadcast herbicide application, with well-documented costs to soil health, biodiversity, and crop physiology. Robotic platforms offer a path toward selective, site-specific control, but orchard environments present challenges such as irregular geometries, trunks, and strong illumination variability under the canopy that have limited fully integrated solutions. This work presents an autonomous robotic platform for selective under-canopy weed control in woody crops, combining multi-sensor perception, a six-degree-of-freedom robotic arm with a mechanical trunk-avoidance mechanism, and a precision spraying module with independently controlled nozzles. A weed image dataset tailored to Mediterranean orchard conditions was built from controlled-cultivation and commercial-orchard imagery under a two-phase training strategy, and the platform was evaluated in a commercial almond orchard in southern Spain. Field trials confirmed the platform’s ability to avoid tree trunks (presenting an average of 1.28% coverage near the tree trunks) and spray only selected targets under typical orchard operation but weed detection accuracy dropped substantially between winter conditions (mAP@0.5 = 93.5%) and summer conditions (mAP@0.5 = 47.8%), with uneven canopy lighting identified as the main cause. These results confirm the technical feasibility of integrating perception, navigation, and actuation into a single autonomous platform, while highlighting robust perception under canopy-induced illumination heterogeneity and tighter perception–navigation integration as the main remaining challenges.
- 11 August 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (16) , 7982
Open AccessIntegration of AI-Based Weed Detection and Robotic Actuation for Site-Specific Under-Canopy Spraying in Woody Crops
Weed management in woody perennial crops relies mainly on broadcast herbicide application, with well-documented costs to soil health, biodiversity, and crop physiology. Robotic platforms offer a path toward selective, site-specific control, but orchard environments present challenges such as irregular geometries, trunks, and strong illumination variability under the canopy that have limited fully integrated solutions. This work presents an autonomous robotic platform for selective under-canopy weed control in woody crops, combining multi-sensor perception, a six-degree-of-freedom robotic arm with a mechanical trunk-avoidance mechanism, and a precision spraying module with independently controlled nozzles. A weed image dataset tailored to Mediterranean orchard conditions was built from controlled-cultivation and commercial-orchard imagery under a two-phase training strategy, and the platform was evaluated in a commercial almond orchard in southern Spain. Field trials confirmed the platform’s ability to avoid tree trunks (presenting an average of 1.28% coverage near the tree trunks) and spray only selected targets under typical orchard operation but weed detection accuracy dropped substantially between winter conditions (mAP@0.5 = 93.5%) and summer conditions (mAP@0.5 = 47.8%), with uneven canopy lighting identified as the main cause. These results confirm the technical feasibility of integrating perception, navigation, and actuation into a single autonomous platform, while highlighting robust perception under canopy-induced illumination heterogeneity and tighter perception–navigation integration as the main remaining challenges.
- Open Access
- 27 April 2026
- Published by Elsevier in SSRN Electronic Journal
Integration of AI-Based Weed Detection and Robotic Actuation for Site-Specific Under-Canopy Spraying in Woody Crops
Weed management in woody perennial crops relies heavily on broadcast herbicide applications, where herbicide is sprayed over the entire area rather than targeting specific weeds. This has well-documented consequences for soil health, biodiversity, and crop physiology. Robotic platforms offer a route toward selective, site-specific control, meaning they can target patches or individual weeds instead of treating the whole field. However, orchard environments present distinct challenges: irregular geometries, trunk obstacles, and strong illumination heterogeneity under the canopy, which have limited the development of fully integrated solutions.This work presents an autonomous robotic platform designed for selective under-canopy weed management in woody crops (named Greenbot). The system combines multi-sensor perception (using several types of sensors for data collection), RTK-GNSS/IMU navigation, LiDAR-based trunk detection, a six-degree-of-freedom robotic arm with a mechanical trunk-avoidance mechanism, and a precision spraying module with independently controlled nozzles. Weed detection relies on a YOLOv8-L model trained using a two-phase strategy: homogeneous-illumination data to establish baseline morphological representations, followed by fine-tuning on real canopy-shadow imagery to improve robustness.Field trials in a commercial almond orchard in southern Spain validated the platform's ability to avoid tree trunks and spray only selected targets under typical orchard operation. However, the system's ability to detect objects showed a substantial drop between laboratory and orchard conditions, with uneven lighting identified as the main cause of poorer detection.The results confirm the technical feasibility of selective, under-canopy robotic weed control in woody crops and outline the technical barriers that must be addressed before such systems can operate autonomously in commercial orchards. Specifically, the main challenges are ensuring robust perception under canopy-induced illumination heterogeneity and achieving tighter integration between perception and navigation to enable fully autonomous selective spraying in orchard environments.
- 27 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessIntegration of AI-Based Weed Detection and Robotic Actuation for Site-Specific Under-Canopy Spraying in Woody Crops
Weed management in woody perennial crops relies heavily on broadcast herbicide applications, where herbicide is sprayed over the entire area rather than targeting specific weeds. This has well-documented consequences for soil health, biodiversity, and crop physiology. Robotic platforms offer a route toward selective, site-specific control, meaning they can target patches or individual weeds instead of treating the whole field. However, orchard environments present distinct challenges: irregular geometries, trunk obstacles, and strong illumination heterogeneity under the canopy, which have limited the development of fully integrated solutions.This work presents an autonomous robotic platform designed for selective under-canopy weed management in woody crops (named Greenbot). The system combines multi-sensor perception (using several types of sensors for data collection), RTK-GNSS/IMU navigation, LiDAR-based trunk detection, a six-degree-of-freedom robotic arm with a mechanical trunk-avoidance mechanism, and a precision spraying module with independently controlled nozzles. Weed detection relies on a YOLOv8-L model trained using a two-phase strategy: homogeneous-illumination data to establish baseline morphological representations, followed by fine-tuning on real canopy-shadow imagery to improve robustness.Field trials in a commercial almond orchard in southern Spain validated the platform's ability to avoid tree trunks and spray only selected targets under typical orchard operation. However, the system's ability to detect objects showed a substantial drop between laboratory and orchard conditions, with uneven lighting identified as the main cause of poorer detection.The results confirm the technical feasibility of selective, under-canopy robotic weed control in woody crops and outline the technical barriers that must be addressed before such systems can operate autonomously in commercial orchards. Specifically, the main challenges are ensuring robust perception under canopy-induced illumination heterogeneity and achieving tighter integration between perception and navigation to enable fully autonomous selective spraying in orchard environments.