Evaluation of Deep Learning Segmentation Models for Detection of Pine Wilt Disease in Unmanned Aerial Vehicle Images

Abstract

Pine wilt disease (PWD) is a serious threat to pine forests. Combining unmanned aerial vehicle (UAV) images and deep learning (DL) techniques to identify infected pines is the most efficient method to determine the potential spread of PWD over a large area. In particular, image segmentation using DL obtains the detailed shape and size of infected pines to assess the disease’s degree of damage. However, the performance of such segmentation models has not been thoroughly studied. We used a fixed-wing UAV to collect images from a pine forest in Laoshan, Qingdao, China, and conducted a ground survey to collect samples of infected pines and construct prior knowledge to interpret the images. Then, training and test sets were annotated on selected images, and we obtained 2352 samples of infected pines annotated over different backgrounds. Finally, high-performance DL models (e.g., fully convolutional networks for semantic segmentation, DeepLabv3+, and PSPNet) were trained and evaluated. The results demonstrated that focal loss provided a higher accuracy and a finer boundary than Dice loss, with the average intersection over union (IoU) for all models increasing from 0.656 to 0.701. From the evaluated models, DeepLLabv3+ achieved the highest IoU and an F1 score of 0.720 and 0.832, respectively. Also, an atrous spatial pyramid pooling module encoded multiscale context information, and the encoder–decoder architecture recovered location/spatial information, being the best architecture for segmenting trees infected by the PWD. Furthermore, segmentation accuracy did not improve as the depth of the backbone network increased, and neither ResNet34 nor ResNet50 was the appropriate backbone for most segmentation models.

UAV-Based Forest Health Monitoring: A Systematic Review

“…RGB cameras capture visible light approximately within the 400-700 nm band of the electromagnetic spectrum. Depending on the sensor and focal length, subdecimeter spatial resolutions can be achieved even from relatively high altitudes above ground when attached to drones [44]. It is common practice in RS to separate color channels to work with the individual bandwidths [45], for example, using different types of filters [46].…”

“…Soil sampling was conducted to determine possible origins of nutritional deficiencies [109] and to measure the field capacity [147]. Ground-based photographs were taken to calculate canopy cover [148], for the documentation of weather conditions [137], and for an improved categorization of pest infestation [44,118,142,146], disease [122], and fire [149] severity classes. Smigaj et al [136] collected data from intratrunk water flow, canopy temperature, soil moisture, and incident and reflected light using an array of sensors.…”

“…Utilizing an RF classifier, they found the highest accuracy by fusing both sensors to distinguish five stages of infection severity (OA: 73.96%, kappa: 0.66). Xia et al [44] instead used unprocessed UAV-based RGB imagery and showed that deep learning approaches represented by several CNN models were superior to the more traditional RF classifier for detecting the pine wilt disease. All papers on diseases have been collected in Appendix A Table A2.…”

“…In some cases, spectral data were enriched with structural information derived from point clouds, DSMs or CHMs based on either LiDAR [116][117][118] or image data [96,102,119,120], further improving classification results. In a few papers, the raw drone images were directly analyzed without further processing [44,108,[121][122][123][124][125][126]. To create photogrammetric products, the researchers predominantly implemented commercial SfM software such as Agisoft Metashape (Agisoft LLC, St. Petersburg, Russia) and Pix4D (Pix4D S.A., Lausanne, Switzerland).…”

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

“…For example, Zhang et al [56] designed a spatiotemporal change detection method in a complex landscape, using deep learning algorithms to capture the spectral, temporal, and spatial characteristics of the target from the image, thereby reducing false detections in tree-scale PWD monitoring. In another study, in order to obtain the detailed shape and size of infected pines, high-performance deep learning models (e.g., fully convolutional networks for semantic segmentation) were applied to perform image segmentation to evaluate the disease's degree of damage, and achieved good results [57].…”