COMPUTER VISION SYSTEM FOR DETECTING ORCHARD TREES FROM UAV IMAGES

Abstract. Orchard tree inventory plays an important role in acquiring up-to-date information on planted trees for effective treatments and crop insurance purposes. Determining damage could help assess orchards’ health faster cheaper. Having accurate the tree’s status also managers to plan necessary fieldwork predict productivity. Traditional orchard is often performed manually, thus time-consuming, costly, subject error. An alternative computer vision algorithms that automatically detect based UAV imagery. The objective of this study develop a method using advanced apple multispectral images. This task challenging since are overlapping over images, hence distinguishing different crowns be difficult. Motivated by latest advances imagery deep-learning models, addressed detection problem exploring two CNN models YOLO (You Only Look Once) DeepForest detecting We first constructed labelled dataset dividing area into equally sized patches. Then we manually annotated all seen RGB was then randomly divided three subsets (training, validation, testing), training testing machine learning models. experiments demonstrate efficiency validity proposed approach inventory. In particular, framework achieved precision 91% F1-score 87% adopting model detection.