Naslov
Detection of pear leaf blister moth using an automatic pest monitoring system

Autori
Miklečić, Ivana ; Čirjak, Dana ; Lemić, Darija ; Kos, Tomislav ; Pajač Živković, Ivana

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Book of Abstracts / Plant Health in Sustainable Agriculture: Hot Spots and Solution Perspectives - Novi Sad, Srbija : Erasmus+ KA2 Program of the European Union / - , 2022, 10-10

Skup
Plant Health in Sustainable Agriculture: Hot Spots and Solution Perspectives

Mjesto i datum
Novi Sad, Srbija, 06.09.2022. - 08.09.2022

Vrsta sudjelovanja
Poster

Vrsta recenzije
Recenziran

Ključne riječi
Leucoptera malifoliella (Costa 1836), apple production, artificial neural network, precision agriculture

Sažetak
Pear leaf blister moth (Lepidoptera: Lionetiidae) is an economically dangerous insect in apple orchards worldwide. During the growing season, the moth develops several generations that cause damage directly to the leaves. The larvae live inside the leaves and feed on the mesophyll tissue, causing defoliation of the leaves and later affecting bud differentiation. In more severe infestations, fruit organoleptic characteristics may also be affected. In the context of precision agriculture, the introduction of new technologies and early detection of pests through the use of automated monitoring systems for economically important apple pests can improve pest control and reduce damage to apple crops. The objective of this study is to develop an automatic monitoring system based on pear leaf blister moth detection using an RGB camera. During a period of 11 weeks, a series of 250 images were taken, and 4150 moths were annotated in the images. A convolutional neural network was trained based on 90% of the annotated images. The results showed that the analytical model trained in this way was able to identify the pear leaf blister moth with 70% accuracy, which was verified using the remaining 10% of the captured images. These preliminary results show that the system can contribute to more accurate and early detection of pests, but still needs to be improved with a larger training data set.

Izvorni jezik
Engleski

Znanstvena područja
Poljoprivreda (agronomija), Informacijske i komunikacijske znanosti

POVEZANOST RADA