Assessment of defoliation during the Dendrolimus tabulaeformis Tsai et Liu disaster outbreak using UAV-based hyperspectral images

The increased frequency and intensity of insect-induced forest disturbances necessitates effective methods to precisely monitor and map the degree of disaster. Unmanned aerial vehicle (UAV)-based hyperspectral imaging is an effective technology for surveying and monitoring forest health. In this article, a novel framework that utilizes a UAV-based hyperspectral image is proposed to identify the degree of damage caused by Dendrolirnus tabulaeformis Tsai et Liu (D.tabulaeformis) in Jianping county of Liaoning province, China. First, data reduction of the hyperspectral image is achieved by comparing three waveband selection algorithms: principal components analysis (PCA), the successive projection algorithm (SPA), and the instability index between classes (ISIC). On this basis, a joint algorithm, ISIC-SPA, which demonstrates the best waveband selection efficiency and good cross-validation accuracy, is proposed. ISIC-SPA is used to select only three sensitive wavebands from 125 original wavebands with a root mean square error of 0.1535. Then, according to analysis of the three sensitive wavebands' reflectance and the corresponding defoliation rate, the piecewise index (PI, B(710 + 738 - 522)) was constructed and the threshold of PI was found to divide the defoliation level. Finally, a piecewise partial least squares regression model was established to quantitatively estimate the defoliation using the optimal wavebands to identify and demarcate the damage level to individual trees. The assessment accuracy of damage caused by D.tabulaeformis at the tree level reached 95.23% using the ISIC-SPA-P-PLSR framework.