Pregled bibliografske jedinice broj: 1264387
Development and Analysis of Models for Detection of Olive Trees
Development and Analysis of Models for Detection of Olive Trees // Advances in science, technology and engineering systems journal, 8 (2023), 87-96 doi:10.25046/aj080210 (međunarodna recenzija, članak, znanstveni)
CROSBI ID: 1264387 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Development and Analysis of Models for Detection of
Olive Trees
Autori
Marin, Ivana ; Gotovac, Sven ; Papić, Vladan
Izvornik
Advances in science, technology and engineering systems journal (2415-6698) 8
(2023);
87-96
Vrsta, podvrsta i kategorija rada
Radovi u časopisima, članak, znanstveni
Ključne riječi
Tree detection ; Olive tree ; Remote sensing ; Deep learning
Sažetak
In this paper, an automatic method for detection of olive trees in RGB images acquired by an unmanned aerial vehicle (UAV) is developed. Presented approach is based on the implementation of RetinaNet model and DeepForest Phyton package. Due to fact that original (pretrained) model used in DeepForest package has been built on images of various types of trees but without images of olive trees, original model detection was unsatisfactory. Therefore, a new image dataset of olive trees was created using sets of images chosen from five olive groves. For neural network training, individual olive trees were manually labeled, and new models were generated. Each model has been trained on different set of images from selected olive groves. Pretrained model and new models were compared and evaluated for various test scenarios. Obtained results showed high precision and recall values of proposed approach and great improvement in performance compared to the pretrained model.
Izvorni jezik
Engleski
