Detection of Bark Beetle Disturbance at Tree Level Using UAS Multispectral Imagery and Deep Learning

Abstract

This study aimed to examine the potential of convolutional neural networks (CNNs) for the detection of individual trees infested by bark beetles in a multispectral high-resolution dataset acquired by an unmanned aerial system (UAS). We compared the performance of three CNN architectures and the random forest (RF) model to classify the trees into four categories: pines, sbbd (longer infested trees when needles turn yellow), sbbg (trees under green attack) and non-infested trees (sh). The best performance was achieved by the Nez4c3b CNN (kappa 0.80) and Safaugu4c3b CNN (kappa 0.76) using only RGB bands. The main misclassifications were between sbbd and sbbg because of the similar spectral responses. Merging sbbd and sbbg into a more general class of infested trees made the selection of model type less important. All tested model types, including RF, were able to detect infested trees with an F-score of the class over 0.90. Nevertheless, the best overall metrics were achieved again by the Safaugu3c3b model (kappa 0.92) and Nez3cb model (kappa 0.87) using only RGB bands. The performance of both models is comparable, but the Nez model has a higher learning rate for this task. Based on our findings, we conclude that the Nez and Safaugu CNN models are superior to the RF models and transfer learning models for the identification of infested trees and for distinguishing between different infestation stages. Therefore, these models can be used not only for basic identification of infested trees but also for monitoring the development of bark beetle disturbance.

Multispectral Imagery Provides Benefits for Mapping Spruce Tree Decline Due to Bark Beetle Infestation When Acquired Late in the Season

“…Recently, UAVs have been used to monitor bark beetle infestations especially in the Czech Republic where millions of cubic meters of trees have been infested due to bark beetles causing significant economic losses [26][27][28][29][30]. Klouček et al [27] used a low-cost UAV-based system with RGB and modified CIR cameras to capture multispectral time series of a bark beetle (I. typographus) outbreak area in the northern part of the Czech Republic.…”

UAV-Based Forest Health Monitoring: A Systematic Review

“…That enables a direct or postprocessed positional correction using real-time kinematic (RTK) or postprocessed kinematic (PPK) solutions. Two papers from 2019 [99,139], two from 2020 [94,140], and five from 2021 [100,112,130,141,142] substantiate this recent trend. Furthermore, other reference sources were used for geometric correction.…”

“…They were also used to assess early stages (e.g., "green attack") of affected trees that cannot clearly be identified by observing the tree crown. For example, trees were examined for signs of early bark beetle infestations (resin flow, entrance holes, and boring dust) [90,120,130] and in doubtful instances also by removing the bark to expose potential egg galleries in the phloem [96].…”

“…Other architectures related to ANNs were multilayer perceptrons [99] and autoencoders [144]. A comparison between different machine learning classifiers was also carried out [119,130,138,142,144,156,157]. The most comprehensive comparative analysis was performed by Wu et al [111].…”

“…The researchers mainly took reference images of single or multiple reflectance targets in the field, with known or directly measured reflectance values across the spectrum of light, typically using the empirical line method (ELM) for calibration during image postprocessing. Occasionally, the target information was enhanced with irradiance values taken from light sensors attached to the UAV [22,94,109,120,123,[129][130][131][132], which were sometimes also used as the only calibration source [87,99,110,133]. In another study, the total incoming irradiance was simulated using data from a sun photometer [85].…”