DETECTION OF TIGUERAS AND TAPES IN COTTON PLANTATIONS: COMPARISON OF ALGORITHMS FOR SELECTIVE SPRAYING
DOI:
https://doi.org/10.56238/arev7n9-141Keywords:
Computer Vision, Selective Spraying, Haar Cascade, YOLOv5, Weeds, CottonAbstract
The presence of plants such as tigueras and ratoons in cotton fields can compromise crop productivity and quality, requiring efficient control strategies, as the boll weevil, one of the largest cotton pests, uses these plants for food and shelter. It is important to eliminate these plants as soon as the cotton is harvested. To achieve this, a system can be used that detects tigueras and ratoons and sprays herbicide directly on these plants, when detected, and only on these plants, reducing the amount of herbicide to be applied at the plantation site. This study compares three computational approaches for detecting these weeds: (1) a heuristic method based on detecting shades of green, (2) an image recognition model using Haar Cascade, and (3) a model based on YOLOv5. The results showed good performance of the Haar Cascade algorithm, which had the highest accuracy rate (89%), followed by YOLOv5 (85%). The heuristic method demonstrated generic detection for green hues, with 100% accuracy.
Downloads
References
BRUNNER, P. et al. A review on computer vision and artificial intelligence techniques for weed detection in agriculture. Smart Agriculture, v. 8, n. 4, p. 235-256, 2021.
KAPLAN, R.; SANTOS, J. H. Deep Learning aplicado à agricultura de precisão: detecção de ervas daninhas. Revista Brasileira de Tecnologia Agrícola, v. 12, n. 2, p. 54-67, 2022
REDMON, J.; FARHADI, A. YOLOv5: You Only Look Once. 2020. Disponível em: https://github.com/ultralytics/yolov5. Acessado em: mar, 2025.
SINGH, K.; LILLESAND, T. Machine learning approaches for weed detection. Precision Agriculture, v. 10, n. 3, p. 193-204, 2019.
VIOLA, P.; JONES, M. Robust real-time object detection. International Journal of Computer Vision, v. 57, n. 2, p. 137–154, 2004.
