A1 Journal article (refereed)
Estimating Tree Health Decline Caused by Ips typographus L. from UAS RGB Images Using a Deep One-Stage Object Detection Neural Network (2022)
Kanerva, H., Honkavaara, E., Näsi, R., Hakala, T., Junttila, S., Karila, K., Koivumäki, N., Alves Oliveira, R., Pelto-Arvo, M., Pölönen, I., Tuviala, J., Östersund, M., & Lyytikäinen-Saarenmaa, P. (2022). Estimating Tree Health Decline Caused by Ips typographus L. from UAS RGB Images Using a Deep One-Stage Object Detection Neural Network. Remote Sensing, 14(24), Article 6257. https://doi.org/10.3390/rs14246257
JYU authors or editors
Publication details
All authors or editors: Kanerva, Heini; Honkavaara, Eija; Näsi, Roope; Hakala, Teemu; Junttila, Samuli; Karila, Kirsi; Koivumäki, Niko; Alves Oliveira, Raquel; Pelto-Arvo, Mikko; Pölönen, Ilkka; et al.
Journal or series: Remote Sensing
eISSN: 2072-4292
Publication year: 2022
Publication date: 10/12/2022
Volume: 14
Issue number: 24
Article number: 6257
Publisher: MDPI
Publication country: Switzerland
Publication language: English
DOI: https://doi.org/10.3390/rs14246257
Publication open access: Openly available
Publication channel open access: Open Access channel
Publication is parallel published (JYX): https://jyx.jyu.fi/handle/123456789/85315
Publication is parallel published: https://helda.helsinki.fi/handle/10138/352172
Abstract
Various biotic and abiotic stresses are causing decline in forest health globally. Presently, one of the major biotic stress agents in Europe is the European spruce bark beetle (Ips typographus L.) which is increasingly causing widespread tree mortality in northern latitudes as a consequence of the warming climate. Remote sensing using unoccupied aerial systems (UAS) together with evolving machine learning techniques provide a powerful tool for fast-response monitoring of forest health. The aim of this study was to investigate the performance of a deep one-stage object detection neural network in the detection of damage by I. typographus in Norway spruce trees using UAS RGB images. A Scaled-YOLOv4 (You Only Look Once) network was implemented and trained for tree health analysis. Datasets for model training were collected during 2013–2020 from three different areas, using four different RGB cameras, and under varying weather conditions. Different model training options were evaluated, including two different symptom rules, different partitions of the dataset, fine-tuning, and hyperparameter optimization. Our study showed that the network was able to detect and classify spruce trees that had visually separable crown symptoms, but it failed to separate spruce trees with stem symptoms and a green crown from healthy spruce trees. For the best model, the overall F-score was 89%, and the F-scores for the healthy, infested, and dead trees were 90%, 79%, and 98%, respectively. The method adapted well to the diverse dataset, and the processing results with different options were consistent. The results indicated that the proposed method could enable implementation of low-cost tools for management of I. typographus outbreaks.
Keywords: forests; Norway spruce; forest damage; insect damage; Ips typographus; monitoring; estimating (statistical methods); remote sensing; unmanned aerial vehicles; machine learning; deep learning; neural networks (information technology)
Free keywords: bark beetle; deep learning; drone; object detection; remote sensing; tree health
Contributing organizations
Related projects
- Autonomous tree health analyzer based on imaging UAV spectrometry
- Pölönen, Ilkka
- Research Council of Finland
Ministry reporting: Yes
VIRTA submission year: 2022
JUFO rating: 1