Zimo Zhou∗ and Xinting Yang∗∗
Pine wilt disease, object detection, unmanned aerial vehicles
The pine wilt, commonly known as pine wood nematode, is a worldwide disease, and it occurs in much of USA, Canada, and Mexico. As it spread rapidly, it was also discovered in Japan, China, Siberia, and France, which was regarded as an introduced pathogen. To manage this disease, the affected trees should be removed timely to prevent transmission. The development on remote sensing techniques brings opportunity to complete object detection on unmanned aerial vehicles (UAV)-captured photos. In this work, we use Feilong UAV to capture images in several counties in Yichang City, Hubei Province. Faster region based convolutional neural networks (R-CNN) network proposed based on region proposal network shows great improvements on detection task, while onestage detection method, such as single shot detection (SSD) and you only look once (YOLO) series, also performs well on this task. We investigated recent one-stage and two-stage object detection algorithm on our drone-captured images. We demonstrate that YOLOv5 has a better performance on our data.
Important Links:
Go Back
