Detection and counting of plants via deep learning using images collected by RPA
DOI:
https://doi.org/10.5039/agraria.v17i2a1353Keywords:
agriculture, production estimation, RGB images, YOLOAbstract
Plant counting and location are essential to provide better control and production estimates in agricultural regions. Techniques based on deep learning have promising results in several application domains, including image analysis collected by RPA. This paper proposes the use of a deep learning model to detect and count plants in RGB images acquired by an unmanned aerial vehicle. The results were obtained via the YOLO model, with validation performed on manually annotated images. The experimental results of the trained model, considering an overlap greater than or equal to 50%, had an average precision of 84.8% and a recall of 89% for images where training and tests were performed in the same field. Experiments were also carried out with the trained model on images from different regions of the training, demonstrating effective results in detecting plants.