The dataset for identifying diseased pine trees affected by pine wood nematode based on an improved YOLOv8 model with drone multispectral imagery

Description

The dataset for identifying diseased pine trees affected by pine wood nematode based on an improved YOLOv8 model with drone multispectral imagery.This dataset includes the base YOLOv8 model code, YOLOv7 model code, Faster R-CNN model code, the improved YOLOv8 base model code, multispectral drone imagery cropped to 600x600 pixels, and the dataset for training.

Steps to reproduce

The image acquisition took place on July 6 and August 25, 2023, using a DJI M300 drone equipped with a Changguang Yuchen MS600 Pro multispectral camera, covering red, green, blue, near-infrared, red edge 720nm, and red edge 750nm bands. Through multiple flights, 3,500 visible light images and 1,000 multispectral images were captured, and sharpening filters were applied to enhance image feature points. A total of 940 stitched images were split into training, validation, and test sets at a ratio of 8:1:1. To enhance model robustness and generalization, the training set underwent data augmentation processes such as rotation, flipping, and scaling, expanding from 750 to 6,050 training images. The 1280x960 pixel images were cropped to 600x600 pixels, and a total of 17,314 images were obtained, with 8,560 for training, 1,070 for validation, and 985 for testing. Tree crowns in the images were annotated using the labelimg tool, creating a YOLO-format dataset. Annotation files were created based on different band features and vegetation indices selected for red edge bands. Experimental observations revealed that PWD (pine wood nematode) progresses slowly in the region, with cases of partial or full crown discoloration. To improve model classification, healthy pines, fully discolored infected trees, and partially discolored infected trees were annotated. Orthophoto images from the drone, combined with GPS and RTK, were used to obtain the coordinates of infected trees. On July 6, needle samples from 15 pine trees were collected and categorized into healthy needles, partially discolored, and fully discolored PWD-infected needles, with 400 samples in each group. Spectral measurements were taken in a darkroom using the ASD Field Spec 4 HR NG spectrometer. Visible light, near-infrared, red edge, NDVI, and NDRE images were combined to train Faster R-CNN, YOLOv8, and YOLOv7 models. Finally, an improved YOLOv8 model was developed to enhance recognition accuracy, supporting large-scale rapid monitoring.All the data/code used for analysis has been made public through Mendeley Data to ensure the reproducibility of the experiment. The training and prediction data used in the deep learning experiment have over 100GB of memory, but due to the limited memory of the Mendeley Data database, only a portion of the data has been uploaded. If anyone needs to replicate my experiment, they can send me a Google email application, and I am happy to provide other database connections for sharing.

Institutions

Categories

Funding

Licence