Three-Dimensional Convolutional Neural Network Model for Early Detection of Pine Wilt Disease Using UAV-Based Hyperspectral Images

Abstract

As one of the most devastating disasters to pine forests, pine wilt disease (PWD) has caused tremendous ecological and economic losses in China. An effective way to prevent large-scale PWD outbreaks is to detect and remove the damaged pine trees at the early stage of PWD infection. However, early infected pine trees do not show obvious changes in morphology or color in the visible wavelength range, making early detection of PWD tricky. Unmanned aerial vehicle (UAV)-based hyperspectral imagery (HI) has great potential for early detection of PWD. However, the commonly used methods, such as the two-dimensional convolutional neural network (2D-CNN), fail to simultaneously extract and fully utilize the spatial and spectral information, whereas the three-dimensional convolutional neural network (3D-CNN) is able to collect this information from raw hyperspectral data. In this paper, we applied the residual block to 3D-CNN and constructed a 3D-Res CNN model, the performance of which was then compared with that of 3D-CNN, 2D-CNN, and 2D-Res CNN in identifying PWD-infected pine trees from the hyperspectral images. The 3D-Res CNN model outperformed the other models, achieving an overall accuracy (OA) of 88.11% and an accuracy of 72.86% for detecting early infected pine trees (EIPs). Using only 20% of the training samples, the OA and EIP accuracy of 3D-Res CNN can still achieve 81.06% and 51.97%, which is superior to the state-of-the-art method in the early detection of PWD based on hyperspectral images. Collectively, 3D-Res CNN was more accurate and effective in early detection of PWD. In conclusion, 3D-Res CNN is proposed for early detection of PWD in this paper, making the prediction and control of PWD more accurate and effective. This model can also be applied to detect pine trees damaged by other diseases or insect pests in the forest.

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.…”

“…Repetitive UAV surveys of the same area were designed to collect time series as a database for change analysis. Hence, the scientists analyzed spectral and structural changes over time to assess mechanical crown damage [97,98], fire damage based on pre-and postfire data [99][100][101], phenological differences [89], and different stages of stress-induced symptoms evident in the tree canopy [92,94,[102][103][104][105][106][107][108][109][110][111][112]. The primary period of FHM typically lay within the growing season.…”

Early detection of pine wilt disease tree candidates using time-series of spectral signatures

“…After each data collection, the data was preprocessed using well-established processing routines ( Holzwarth et al., 2011 ; Einzmann et al., 2021 ; Yu et al., 2021a ), leading to top-of-canopy spectral bi-directional reflectances. The irradiance calibration, reflectance correction, and image mosaicking were conducted using the 3 m 2 standard board and Spectronon software (Resonon, Bozeman, MA, USA), Megacube (LICA United Technology Limited, Beijing, China), ArcGIS (ESRI, Redlands, CA, USA), IDL 8.5 and ENVI 5.3 (Harris Corporation, Melbourne, FL, USA).…”

“…Hyperspectral RS uses data from hundreds of bands and continuous wavelengths, and these bands can capture physiological changes in infested trees, which help detect early-stage pest and diseases infestations ( Cheng et al., 2010 ; Abdullah et al., 2018 ; Liu et al., 2021 ; Yu et al., 2021d ). As the advanced technology used in forest pest and diseases detection, UAV-based hyperspectral imagery (HSI) can provide highly accurate detection with flexible and efficient data acquisition ( Iordache et al., 2020 ; Li et al., 2020 ; Zhang et al., 2018 ; Lin et al., 2019 ; Lin et al., 2021 ; Yu et al., 2021a ). Most studies were based on single-date UAV-based hyperspectral data for early monitoring of forest pest and diseases ( Lausch et al., 2013 ; Li et al., 2020 ; Zhang et al., 2018 ; Lin et al., 2019 ; Lin et al., 2021 ; Yu et al., 2021a ; Yu et al., 2021c ; Yu et al., 2021d ).…”

“…As the advanced technology used in forest pest and diseases detection, UAV-based hyperspectral imagery (HSI) can provide highly accurate detection with flexible and efficient data acquisition ( Iordache et al., 2020 ; Li et al., 2020 ; Zhang et al., 2018 ; Lin et al., 2019 ; Lin et al., 2021 ; Yu et al., 2021a ). Most studies were based on single-date UAV-based hyperspectral data for early monitoring of forest pest and diseases ( Lausch et al., 2013 ; Li et al., 2020 ; Zhang et al., 2018 ; Lin et al., 2019 ; Lin et al., 2021 ; Yu et al., 2021a ; Yu et al., 2021c ; Yu et al., 2021d ). Compared with single-date data, the use of time-series data is relatively rare ( Iordache et al., 2020 ; Einzmann et al., 2021 ; Bárta et al., 2022 ), even though it could be used to capture more complete infestation processes and determine the optimal detection periods, thus leading to more reliable results than those based on a single date, with the limitation that any changes observed in the color and texture of tree crowns could also be due to phenology and other factors.…”